net/core/rtnetlink.c

   1 // SPDX-License-Identifier: GPL-2.0-or-later
   2 /*
   3  * INET         An implementation of the TCP/IP protocol suite for the LINUX
   4  *              operating system.  INET is implemented using the  BSD Socket
   5  *              interface as the means of communication with the user level.
   6  *
   7  *              Routing netlink socket interface: protocol independent part.
   8  *
   9  * Authors:     Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
  10  *
  11  *      Fixes:
  12  *      Vitaly E. Lavrov                RTA_OK arithmetics was wrong.
  13  */
  14
  15 #include <linux/bitops.h>
  16 #include <linux/errno.h>
  17 #include <linux/module.h>
  18 #include <linux/types.h>
  19 #include <linux/socket.h>
  20 #include <linux/kernel.h>
  21 #include <linux/timer.h>
  22 #include <linux/string.h>
  23 #include <linux/sockios.h>
  24 #include <linux/net.h>
  25 #include <linux/fcntl.h>
  26 #include <linux/mm.h>
  27 #include <linux/slab.h>
  28 #include <linux/interrupt.h>
  29 #include <linux/capability.h>
  30 #include <linux/skbuff.h>
  31 #include <linux/init.h>
  32 #include <linux/security.h>
  33 #include <linux/mutex.h>
  34 #include <linux/if_addr.h>
  35 #include <linux/if_bridge.h>
  36 #include <linux/if_vlan.h>
  37 #include <linux/pci.h>
  38 #include <linux/etherdevice.h>
  39 #include <linux/bpf.h>
  40
  41 #include <linux/uaccess.h>
  42
  43 #include <linux/inet.h>
  44 #include <linux/netdevice.h>
  45 #include <net/ip.h>
  46 #include <net/protocol.h>
  47 #include <net/arp.h>
  48 #include <net/route.h>
  49 #include <net/udp.h>
  50 #include <net/tcp.h>
  51 #include <net/sock.h>
  52 #include <net/pkt_sched.h>
  53 #include <net/fib_rules.h>
  54 #include <net/rtnetlink.h>
  55 #include <net/net_namespace.h>
  56
  57 #define RTNL_MAX_TYPE           50
  58 #define RTNL_SLAVE_MAX_TYPE     36
  59
  60 struct rtnl_link {
  61         rtnl_doit_func          doit;
  62         rtnl_dumpit_func        dumpit;
  63         struct module           *owner;
  64         unsigned int            flags;
  65         struct rcu_head         rcu;
  66 };
  67
  68 static DEFINE_MUTEX(rtnl_mutex);
  69
  70 void rtnl_lock(void)
  71 {
  72         mutex_lock(&rtnl_mutex);
  73 }
  74 EXPORT_SYMBOL(rtnl_lock);
  75
  76 int rtnl_lock_killable(void)
  77 {
  78         return mutex_lock_killable(&rtnl_mutex);
  79 }
  80 EXPORT_SYMBOL(rtnl_lock_killable);
  81
  82 static struct sk_buff *defer_kfree_skb_list;
  83 void rtnl_kfree_skbs(struct sk_buff *head, struct sk_buff *tail)
  84 {
  85         if (head && tail) {
  86                 tail->next = defer_kfree_skb_list;
  87                 defer_kfree_skb_list = head;
  88         }
  89 }
  90 EXPORT_SYMBOL(rtnl_kfree_skbs);
  91
  92 void __rtnl_unlock(void)
  93 {
  94         struct sk_buff *head = defer_kfree_skb_list;
  95
  96         defer_kfree_skb_list = NULL;
  97
  98         mutex_unlock(&rtnl_mutex);
  99
 100         while (head) {
 101                 struct sk_buff *next = head->next;
 102
 103                 kfree_skb(head);
 104                 cond_resched();
 105                 head = next;
 106         }
 107 }
 108
 109 void rtnl_unlock(void)
 110 {
 111         /* This fellow will unlock it for us. */
 112         netdev_run_todo();
 113 }
 114 EXPORT_SYMBOL(rtnl_unlock);
 115
 116 int rtnl_trylock(void)
 117 {
 118         return mutex_trylock(&rtnl_mutex);
 119 }
 120 EXPORT_SYMBOL(rtnl_trylock);
 121
 122 int rtnl_is_locked(void)
 123 {
 124         return mutex_is_locked(&rtnl_mutex);
 125 }
 126 EXPORT_SYMBOL(rtnl_is_locked);
 127
 128 bool refcount_dec_and_rtnl_lock(refcount_t *r)
 129 {
 130         return refcount_dec_and_mutex_lock(r, &rtnl_mutex);
 131 }
 132 EXPORT_SYMBOL(refcount_dec_and_rtnl_lock);
 133
 134 #ifdef CONFIG_PROVE_LOCKING
 135 bool lockdep_rtnl_is_held(void)
 136 {
 137         return lockdep_is_held(&rtnl_mutex);
 138 }
 139 EXPORT_SYMBOL(lockdep_rtnl_is_held);
 140 #endif /* #ifdef CONFIG_PROVE_LOCKING */
 141
 142 static struct rtnl_link *__rcu *rtnl_msg_handlers[RTNL_FAMILY_MAX + 1];
 143
 144 static inline int rtm_msgindex(int msgtype)
 145 {
 146         int msgindex = msgtype - RTM_BASE;
 147
 148         /*
 149          * msgindex < 0 implies someone tried to register a netlink
 150          * control code. msgindex >= RTM_NR_MSGTYPES may indicate that
 151          * the message type has not been added to linux/rtnetlink.h
 152          */
 153         BUG_ON(msgindex < 0 || msgindex >= RTM_NR_MSGTYPES);
 154
 155         return msgindex;
 156 }
 157
 158 static struct rtnl_link *rtnl_get_link(int protocol, int msgtype)
 159 {
 160         struct rtnl_link **tab;
 161
 162         if (protocol >= ARRAY_SIZE(rtnl_msg_handlers))
 163                 protocol = PF_UNSPEC;
 164
 165         tab = rcu_dereference_rtnl(rtnl_msg_handlers[protocol]);
 166         if (!tab)
 167                 tab = rcu_dereference_rtnl(rtnl_msg_handlers[PF_UNSPEC]);
 168
 169         return tab[msgtype];
 170 }
 171
 172 static int rtnl_register_internal(struct module *owner,
 173                                   int protocol, int msgtype,
 174                                   rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 175                                   unsigned int flags)
 176 {
 177         struct rtnl_link *link, *old;
 178         struct rtnl_link __rcu **tab;
 179         int msgindex;
 180         int ret = -ENOBUFS;
 181
 182         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 183         msgindex = rtm_msgindex(msgtype);
 184
 185         rtnl_lock();
 186         tab = rtnl_msg_handlers[protocol];
 187         if (tab == NULL) {
 188                 tab = kcalloc(RTM_NR_MSGTYPES, sizeof(void *), GFP_KERNEL);
 189                 if (!tab)
 190                         goto unlock;
 191
 192                 /* ensures we see the 0 stores */
 193                 rcu_assign_pointer(rtnl_msg_handlers[protocol], tab);
 194         }
 195
 196         old = rtnl_dereference(tab[msgindex]);
 197         if (old) {
 198                 link = kmemdup(old, sizeof(*old), GFP_KERNEL);
 199                 if (!link)
 200                         goto unlock;
 201         } else {
 202                 link = kzalloc(sizeof(*link), GFP_KERNEL);
 203                 if (!link)
 204                         goto unlock;
 205         }
 206
 207         WARN_ON(link->owner && link->owner != owner);
 208         link->owner = owner;
 209
 210         WARN_ON(doit && link->doit && link->doit != doit);
 211         if (doit)
 212                 link->doit = doit;
 213         WARN_ON(dumpit && link->dumpit && link->dumpit != dumpit);
 214         if (dumpit)
 215                 link->dumpit = dumpit;
 216
 217         link->flags |= flags;
 218
 219         /* publish protocol:msgtype */
 220         rcu_assign_pointer(tab[msgindex], link);
 221         ret = 0;
 222         if (old)
 223                 kfree_rcu(old, rcu);
 224 unlock:
 225         rtnl_unlock();
 226         return ret;
 227 }
 228
 229 /**
 230  * rtnl_register_module - Register a rtnetlink message type
 231  *
 232  * @owner: module registering the hook (THIS_MODULE)
 233  * @protocol: Protocol family or PF_UNSPEC
 234  * @msgtype: rtnetlink message type
 235  * @doit: Function pointer called for each request message
 236  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
 237  * @flags: rtnl_link_flags to modifiy behaviour of doit/dumpit functions
 238  *
 239  * Like rtnl_register, but for use by removable modules.
 240  */
 241 int rtnl_register_module(struct module *owner,
 242                          int protocol, int msgtype,
 243                          rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 244                          unsigned int flags)
 245 {
 246         return rtnl_register_internal(owner, protocol, msgtype,
 247                                       doit, dumpit, flags);
 248 }
 249 EXPORT_SYMBOL_GPL(rtnl_register_module);
 250
 251 /**
 252  * rtnl_register - Register a rtnetlink message type
 253  * @protocol: Protocol family or PF_UNSPEC
 254  * @msgtype: rtnetlink message type
 255  * @doit: Function pointer called for each request message
 256  * @dumpit: Function pointer called for each dump request (NLM_F_DUMP) message
 257  * @flags: rtnl_link_flags to modifiy behaviour of doit/dumpit functions
 258  *
 259  * Registers the specified function pointers (at least one of them has
 260  * to be non-NULL) to be called whenever a request message for the
 261  * specified protocol family and message type is received.
 262  *
 263  * The special protocol family PF_UNSPEC may be used to define fallback
 264  * function pointers for the case when no entry for the specific protocol
 265  * family exists.
 266  */
 267 void rtnl_register(int protocol, int msgtype,
 268                    rtnl_doit_func doit, rtnl_dumpit_func dumpit,
 269                    unsigned int flags)
 270 {
 271         int err;
 272
 273         err = rtnl_register_internal(NULL, protocol, msgtype, doit, dumpit,
 274                                      flags);
 275         if (err)
 276                 pr_err("Unable to register rtnetlink message handler, "
 277                        "protocol = %d, message type = %d\n", protocol, msgtype);
 278 }
 279
 280 /**
 281  * rtnl_unregister - Unregister a rtnetlink message type
 282  * @protocol: Protocol family or PF_UNSPEC
 283  * @msgtype: rtnetlink message type
 284  *
 285  * Returns 0 on success or a negative error code.
 286  */
 287 int rtnl_unregister(int protocol, int msgtype)
 288 {
 289         struct rtnl_link **tab, *link;
 290         int msgindex;
 291
 292         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 293         msgindex = rtm_msgindex(msgtype);
 294
 295         rtnl_lock();
 296         tab = rtnl_dereference(rtnl_msg_handlers[protocol]);
 297         if (!tab) {
 298                 rtnl_unlock();
 299                 return -ENOENT;
 300         }
 301
 302         link = tab[msgindex];
 303         rcu_assign_pointer(tab[msgindex], NULL);
 304         rtnl_unlock();
 305
 306         kfree_rcu(link, rcu);
 307
 308         return 0;
 309 }
 310 EXPORT_SYMBOL_GPL(rtnl_unregister);
 311
 312 /**
 313  * rtnl_unregister_all - Unregister all rtnetlink message type of a protocol
 314  * @protocol : Protocol family or PF_UNSPEC
 315  *
 316  * Identical to calling rtnl_unregster() for all registered message types
 317  * of a certain protocol family.
 318  */
 319 void rtnl_unregister_all(int protocol)
 320 {
 321         struct rtnl_link **tab, *link;
 322         int msgindex;
 323
 324         BUG_ON(protocol < 0 || protocol > RTNL_FAMILY_MAX);
 325
 326         rtnl_lock();
 327         tab = rtnl_msg_handlers[protocol];
 328         if (!tab) {
 329                 rtnl_unlock();
 330                 return;
 331         }
 332         RCU_INIT_POINTER(rtnl_msg_handlers[protocol], NULL);
 333         for (msgindex = 0; msgindex < RTM_NR_MSGTYPES; msgindex++) {
 334                 link = tab[msgindex];
 335                 if (!link)
 336                         continue;
 337
 338                 rcu_assign_pointer(tab[msgindex], NULL);
 339                 kfree_rcu(link, rcu);
 340         }
 341         rtnl_unlock();
 342
 343         synchronize_net();
 344
 345         kfree(tab);
 346 }
 347 EXPORT_SYMBOL_GPL(rtnl_unregister_all);
 348
 349 static LIST_HEAD(link_ops);
 350
 351 static const struct rtnl_link_ops *rtnl_link_ops_get(const char *kind)
 352 {
 353         const struct rtnl_link_ops *ops;
 354
 355         list_for_each_entry(ops, &link_ops, list) {
 356                 if (!strcmp(ops->kind, kind))
 357                         return ops;
 358         }
 359         return NULL;
 360 }
 361
 362 /**
 363  * __rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 364  * @ops: struct rtnl_link_ops * to register
 365  *
 366  * The caller must hold the rtnl_mutex. This function should be used
 367  * by drivers that create devices during module initialization. It
 368  * must be called before registering the devices.
 369  *
 370  * Returns 0 on success or a negative error code.
 371  */
 372 int __rtnl_link_register(struct rtnl_link_ops *ops)
 373 {
 374         if (rtnl_link_ops_get(ops->kind))
 375                 return -EEXIST;
 376
 377         /* The check for setup is here because if ops
 378          * does not have that filled up, it is not possible
 379          * to use the ops for creating device. So do not
 380          * fill up dellink as well. That disables rtnl_dellink.
 381          */
 382         if (ops->setup && !ops->dellink)
 383                 ops->dellink = unregister_netdevice_queue;
 384
 385         list_add_tail(&ops->list, &link_ops);
 386         return 0;
 387 }
 388 EXPORT_SYMBOL_GPL(__rtnl_link_register);
 389
 390 /**
 391  * rtnl_link_register - Register rtnl_link_ops with rtnetlink.
 392  * @ops: struct rtnl_link_ops * to register
 393  *
 394  * Returns 0 on success or a negative error code.
 395  */
 396 int rtnl_link_register(struct rtnl_link_ops *ops)
 397 {
 398         int err;
 399
 400         /* Sanity-check max sizes to avoid stack buffer overflow. */
 401         if (WARN_ON(ops->maxtype > RTNL_MAX_TYPE ||
 402                     ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE))
 403                 return -EINVAL;
 404
 405         rtnl_lock();
 406         err = __rtnl_link_register(ops);
 407         rtnl_unlock();
 408         return err;
 409 }
 410 EXPORT_SYMBOL_GPL(rtnl_link_register);
 411
 412 static void __rtnl_kill_links(struct net *net, struct rtnl_link_ops *ops)
 413 {
 414         struct net_device *dev;
 415         LIST_HEAD(list_kill);
 416
 417         for_each_netdev(net, dev) {
 418                 if (dev->rtnl_link_ops == ops)
 419                         ops->dellink(dev, &list_kill);
 420         }
 421         unregister_netdevice_many(&list_kill);
 422 }
 423
 424 /**
 425  * __rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 426  * @ops: struct rtnl_link_ops * to unregister
 427  *
 428  * The caller must hold the rtnl_mutex and guarantee net_namespace_list
 429  * integrity (hold pernet_ops_rwsem for writing to close the race
 430  * with setup_net() and cleanup_net()).
 431  */
 432 void __rtnl_link_unregister(struct rtnl_link_ops *ops)
 433 {
 434         struct net *net;
 435
 436         for_each_net(net) {
 437                 __rtnl_kill_links(net, ops);
 438         }
 439         list_del(&ops->list);
 440 }
 441 EXPORT_SYMBOL_GPL(__rtnl_link_unregister);
 442
 443 /* Return with the rtnl_lock held when there are no network
 444  * devices unregistering in any network namespace.
 445  */
 446 static void rtnl_lock_unregistering_all(void)
 447 {
 448         struct net *net;
 449         bool unregistering;
 450         DEFINE_WAIT_FUNC(wait, woken_wake_function);
 451
 452         add_wait_queue(&netdev_unregistering_wq, &wait);
 453         for (;;) {
 454                 unregistering = false;
 455                 rtnl_lock();
 456                 /* We held write locked pernet_ops_rwsem, and parallel
 457                  * setup_net() and cleanup_net() are not possible.
 458                  */
 459                 for_each_net(net) {
 460                         if (net->dev_unreg_count > 0) {
 461                                 unregistering = true;
 462                                 break;
 463                         }
 464                 }
 465                 if (!unregistering)
 466                         break;
 467                 __rtnl_unlock();
 468
 469                 wait_woken(&wait, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
 470         }
 471         remove_wait_queue(&netdev_unregistering_wq, &wait);
 472 }
 473
 474 /**
 475  * rtnl_link_unregister - Unregister rtnl_link_ops from rtnetlink.
 476  * @ops: struct rtnl_link_ops * to unregister
 477  */
 478 void rtnl_link_unregister(struct rtnl_link_ops *ops)
 479 {
 480         /* Close the race with setup_net() and cleanup_net() */
 481         down_write(&pernet_ops_rwsem);
 482         rtnl_lock_unregistering_all();
 483         __rtnl_link_unregister(ops);
 484         rtnl_unlock();
 485         up_write(&pernet_ops_rwsem);
 486 }
 487 EXPORT_SYMBOL_GPL(rtnl_link_unregister);
 488
 489 static size_t rtnl_link_get_slave_info_data_size(const struct net_device *dev)
 490 {
 491         struct net_device *master_dev;
 492         const struct rtnl_link_ops *ops;
 493         size_t size = 0;
 494
 495         rcu_read_lock();
 496
 497         master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
 498         if (!master_dev)
 499                 goto out;
 500
 501         ops = master_dev->rtnl_link_ops;
 502         if (!ops || !ops->get_slave_size)
 503                 goto out;
 504         /* IFLA_INFO_SLAVE_DATA + nested data */
 505         size = nla_total_size(sizeof(struct nlattr)) +
 506                ops->get_slave_size(master_dev, dev);
 507
 508 out:
 509         rcu_read_unlock();
 510         return size;
 511 }
 512
 513 static size_t rtnl_link_get_size(const struct net_device *dev)
 514 {
 515         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 516         size_t size;
 517
 518         if (!ops)
 519                 return 0;
 520
 521         size = nla_total_size(sizeof(struct nlattr)) + /* IFLA_LINKINFO */
 522                nla_total_size(strlen(ops->kind) + 1);  /* IFLA_INFO_KIND */
 523
 524         if (ops->get_size)
 525                 /* IFLA_INFO_DATA + nested data */
 526                 size += nla_total_size(sizeof(struct nlattr)) +
 527                         ops->get_size(dev);
 528
 529         if (ops->get_xstats_size)
 530                 /* IFLA_INFO_XSTATS */
 531                 size += nla_total_size(ops->get_xstats_size(dev));
 532
 533         size += rtnl_link_get_slave_info_data_size(dev);
 534
 535         return size;
 536 }
 537
 538 static LIST_HEAD(rtnl_af_ops);
 539
 540 static const struct rtnl_af_ops *rtnl_af_lookup(const int family)
 541 {
 542         const struct rtnl_af_ops *ops;
 543
 544         list_for_each_entry_rcu(ops, &rtnl_af_ops, list) {
 545                 if (ops->family == family)
 546                         return ops;
 547         }
 548
 549         return NULL;
 550 }
 551
 552 /**
 553  * rtnl_af_register - Register rtnl_af_ops with rtnetlink.
 554  * @ops: struct rtnl_af_ops * to register
 555  *
 556  * Returns 0 on success or a negative error code.
 557  */
 558 void rtnl_af_register(struct rtnl_af_ops *ops)
 559 {
 560         rtnl_lock();
 561         list_add_tail_rcu(&ops->list, &rtnl_af_ops);
 562         rtnl_unlock();
 563 }
 564 EXPORT_SYMBOL_GPL(rtnl_af_register);
 565
 566 /**
 567  * rtnl_af_unregister - Unregister rtnl_af_ops from rtnetlink.
 568  * @ops: struct rtnl_af_ops * to unregister
 569  */
 570 void rtnl_af_unregister(struct rtnl_af_ops *ops)
 571 {
 572         rtnl_lock();
 573         list_del_rcu(&ops->list);
 574         rtnl_unlock();
 575
 576         synchronize_rcu();
 577 }
 578 EXPORT_SYMBOL_GPL(rtnl_af_unregister);
 579
 580 static size_t rtnl_link_get_af_size(const struct net_device *dev,
 581                                     u32 ext_filter_mask)
 582 {
 583         struct rtnl_af_ops *af_ops;
 584         size_t size;
 585
 586         /* IFLA_AF_SPEC */
 587         size = nla_total_size(sizeof(struct nlattr));
 588
 589         rcu_read_lock();
 590         list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
 591                 if (af_ops->get_link_af_size) {
 592                         /* AF_* + nested data */
 593                         size += nla_total_size(sizeof(struct nlattr)) +
 594                                 af_ops->get_link_af_size(dev, ext_filter_mask);
 595                 }
 596         }
 597         rcu_read_unlock();
 598
 599         return size;
 600 }
 601
 602 static bool rtnl_have_link_slave_info(const struct net_device *dev)
 603 {
 604         struct net_device *master_dev;
 605         bool ret = false;
 606
 607         rcu_read_lock();
 608
 609         master_dev = netdev_master_upper_dev_get_rcu((struct net_device *)dev);
 610         if (master_dev && master_dev->rtnl_link_ops)
 611                 ret = true;
 612         rcu_read_unlock();
 613         return ret;
 614 }
 615
 616 static int rtnl_link_slave_info_fill(struct sk_buff *skb,
 617                                      const struct net_device *dev)
 618 {
 619         struct net_device *master_dev;
 620         const struct rtnl_link_ops *ops;
 621         struct nlattr *slave_data;
 622         int err;
 623
 624         master_dev = netdev_master_upper_dev_get((struct net_device *) dev);
 625         if (!master_dev)
 626                 return 0;
 627         ops = master_dev->rtnl_link_ops;
 628         if (!ops)
 629                 return 0;
 630         if (nla_put_string(skb, IFLA_INFO_SLAVE_KIND, ops->kind) < 0)
 631                 return -EMSGSIZE;
 632         if (ops->fill_slave_info) {
 633                 slave_data = nla_nest_start_noflag(skb, IFLA_INFO_SLAVE_DATA);
 634                 if (!slave_data)
 635                         return -EMSGSIZE;
 636                 err = ops->fill_slave_info(skb, master_dev, dev);
 637                 if (err < 0)
 638                         goto err_cancel_slave_data;
 639                 nla_nest_end(skb, slave_data);
 640         }
 641         return 0;
 642
 643 err_cancel_slave_data:
 644         nla_nest_cancel(skb, slave_data);
 645         return err;
 646 }
 647
 648 static int rtnl_link_info_fill(struct sk_buff *skb,
 649                                const struct net_device *dev)
 650 {
 651         const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
 652         struct nlattr *data;
 653         int err;
 654
 655         if (!ops)
 656                 return 0;
 657         if (nla_put_string(skb, IFLA_INFO_KIND, ops->kind) < 0)
 658                 return -EMSGSIZE;
 659         if (ops->fill_xstats) {
 660                 err = ops->fill_xstats(skb, dev);
 661                 if (err < 0)
 662                         return err;
 663         }
 664         if (ops->fill_info) {
 665                 data = nla_nest_start_noflag(skb, IFLA_INFO_DATA);
 666                 if (data == NULL)
 667                         return -EMSGSIZE;
 668                 err = ops->fill_info(skb, dev);
 669                 if (err < 0)
 670                         goto err_cancel_data;
 671                 nla_nest_end(skb, data);
 672         }
 673         return 0;
 674
 675 err_cancel_data:
 676         nla_nest_cancel(skb, data);
 677         return err;
 678 }
 679
 680 static int rtnl_link_fill(struct sk_buff *skb, const struct net_device *dev)
 681 {
 682         struct nlattr *linkinfo;
 683         int err = -EMSGSIZE;
 684
 685         linkinfo = nla_nest_start_noflag(skb, IFLA_LINKINFO);
 686         if (linkinfo == NULL)
 687                 goto out;
 688
 689         err = rtnl_link_info_fill(skb, dev);
 690         if (err < 0)
 691                 goto err_cancel_link;
 692
 693         err = rtnl_link_slave_info_fill(skb, dev);
 694         if (err < 0)
 695                 goto err_cancel_link;
 696
 697         nla_nest_end(skb, linkinfo);
 698         return 0;
 699
 700 err_cancel_link:
 701         nla_nest_cancel(skb, linkinfo);
 702 out:
 703         return err;
 704 }
 705
 706 int rtnetlink_send(struct sk_buff *skb, struct net *net, u32 pid, unsigned int group, int echo)
 707 {
 708         struct sock *rtnl = net->rtnl;
 709         int err = 0;
 710
 711         NETLINK_CB(skb).dst_group = group;
 712         if (echo)
 713                 refcount_inc(&skb->users);
 714         netlink_broadcast(rtnl, skb, pid, group, GFP_KERNEL);
 715         if (echo)
 716                 err = netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
 717         return err;
 718 }
 719
 720 int rtnl_unicast(struct sk_buff *skb, struct net *net, u32 pid)
 721 {
 722         struct sock *rtnl = net->rtnl;
 723
 724         return nlmsg_unicast(rtnl, skb, pid);
 725 }
 726 EXPORT_SYMBOL(rtnl_unicast);
 727
 728 void rtnl_notify(struct sk_buff *skb, struct net *net, u32 pid, u32 group,
 729                  struct nlmsghdr *nlh, gfp_t flags)
 730 {
 731         struct sock *rtnl = net->rtnl;
 732         int report = 0;
 733
 734         if (nlh)
 735                 report = nlmsg_report(nlh);
 736
 737         nlmsg_notify(rtnl, skb, pid, group, report, flags);
 738 }
 739 EXPORT_SYMBOL(rtnl_notify);
 740
 741 void rtnl_set_sk_err(struct net *net, u32 group, int error)
 742 {
 743         struct sock *rtnl = net->rtnl;
 744
 745         netlink_set_err(rtnl, 0, group, error);
 746 }
 747 EXPORT_SYMBOL(rtnl_set_sk_err);
 748
 749 int rtnetlink_put_metrics(struct sk_buff *skb, u32 *metrics)
 750 {
 751         struct nlattr *mx;
 752         int i, valid = 0;
 753
 754         /* nothing is dumped for dst_default_metrics, so just skip the loop */
 755         if (metrics == dst_default_metrics.metrics)
 756                 return 0;
 757
 758         mx = nla_nest_start_noflag(skb, RTA_METRICS);
 759         if (mx == NULL)
 760                 return -ENOBUFS;
 761
 762         for (i = 0; i < RTAX_MAX; i++) {
 763                 if (metrics[i]) {
 764                         if (i == RTAX_CC_ALGO - 1) {
 765                                 char tmp[TCP_CA_NAME_MAX], *name;
 766
 767                                 name = tcp_ca_get_name_by_key(metrics[i], tmp);
 768                                 if (!name)
 769                                         continue;
 770                                 if (nla_put_string(skb, i + 1, name))
 771                                         goto nla_put_failure;
 772                         } else if (i == RTAX_FEATURES - 1) {
 773                                 u32 user_features = metrics[i] & RTAX_FEATURE_MASK;
 774
 775                                 if (!user_features)
 776                                         continue;
 777                                 BUILD_BUG_ON(RTAX_FEATURE_MASK & DST_FEATURE_MASK);
 778                                 if (nla_put_u32(skb, i + 1, user_features))
 779                                         goto nla_put_failure;
 780                         } else {
 781                                 if (nla_put_u32(skb, i + 1, metrics[i]))
 782                                         goto nla_put_failure;
 783                         }
 784                         valid++;
 785                 }
 786         }
 787
 788         if (!valid) {
 789                 nla_nest_cancel(skb, mx);
 790                 return 0;
 791         }
 792
 793         return nla_nest_end(skb, mx);
 794
 795 nla_put_failure:
 796         nla_nest_cancel(skb, mx);
 797         return -EMSGSIZE;
 798 }
 799 EXPORT_SYMBOL(rtnetlink_put_metrics);
 800
 801 int rtnl_put_cacheinfo(struct sk_buff *skb, struct dst_entry *dst, u32 id,
 802                        long expires, u32 error)
 803 {
 804         struct rta_cacheinfo ci = {
 805                 .rta_error = error,
 806                 .rta_id =  id,
 807         };
 808
 809         if (dst) {
 810                 ci.rta_lastuse = jiffies_delta_to_clock_t(jiffies - dst->lastuse);
 811                 ci.rta_used = dst->__use;
 812                 ci.rta_clntref = atomic_read(&dst->__refcnt);
 813         }
 814         if (expires) {
 815                 unsigned long clock;
 816
 817                 clock = jiffies_to_clock_t(abs(expires));
 818                 clock = min_t(unsigned long, clock, INT_MAX);
 819                 ci.rta_expires = (expires > 0) ? clock : -clock;
 820         }
 821         return nla_put(skb, RTA_CACHEINFO, sizeof(ci), &ci);
 822 }
 823 EXPORT_SYMBOL_GPL(rtnl_put_cacheinfo);
 824
 825 static void set_operstate(struct net_device *dev, unsigned char transition)
 826 {
 827         unsigned char operstate = dev->operstate;
 828
 829         switch (transition) {
 830         case IF_OPER_UP:
 831                 if ((operstate == IF_OPER_DORMANT ||
 832                      operstate == IF_OPER_TESTING ||
 833                      operstate == IF_OPER_UNKNOWN) &&
 834                     !netif_dormant(dev) && !netif_testing(dev))
 835                         operstate = IF_OPER_UP;
 836                 break;
 837
 838         case IF_OPER_TESTING:
 839                 if (operstate == IF_OPER_UP ||
 840                     operstate == IF_OPER_UNKNOWN)
 841                         operstate = IF_OPER_TESTING;
 842                 break;
 843
 844         case IF_OPER_DORMANT:
 845                 if (operstate == IF_OPER_UP ||
 846                     operstate == IF_OPER_UNKNOWN)
 847                         operstate = IF_OPER_DORMANT;
 848                 break;
 849         }
 850
 851         if (dev->operstate != operstate) {
 852                 write_lock_bh(&dev_base_lock);
 853                 dev->operstate = operstate;
 854                 write_unlock_bh(&dev_base_lock);
 855                 netdev_state_change(dev);
 856         }
 857 }
 858
 859 static unsigned int rtnl_dev_get_flags(const struct net_device *dev)
 860 {
 861         return (dev->flags & ~(IFF_PROMISC | IFF_ALLMULTI)) |
 862                (dev->gflags & (IFF_PROMISC | IFF_ALLMULTI));
 863 }
 864
 865 static unsigned int rtnl_dev_combine_flags(const struct net_device *dev,
 866                                            const struct ifinfomsg *ifm)
 867 {
 868         unsigned int flags = ifm->ifi_flags;
 869
 870         /* bugwards compatibility: ifi_change == 0 is treated as ~0 */
 871         if (ifm->ifi_change)
 872                 flags = (flags & ifm->ifi_change) |
 873                         (rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
 874
 875         return flags;
 876 }
 877
 878 static void copy_rtnl_link_stats(struct rtnl_link_stats *a,
 879                                  const struct rtnl_link_stats64 *b)
 880 {
 881         a->rx_packets = b->rx_packets;
 882         a->tx_packets = b->tx_packets;
 883         a->rx_bytes = b->rx_bytes;
 884         a->tx_bytes = b->tx_bytes;
 885         a->rx_errors = b->rx_errors;
 886         a->tx_errors = b->tx_errors;
 887         a->rx_dropped = b->rx_dropped;
 888         a->tx_dropped = b->tx_dropped;
 889
 890         a->multicast = b->multicast;
 891         a->collisions = b->collisions;
 892
 893         a->rx_length_errors = b->rx_length_errors;
 894         a->rx_over_errors = b->rx_over_errors;
 895         a->rx_crc_errors = b->rx_crc_errors;
 896         a->rx_frame_errors = b->rx_frame_errors;
 897         a->rx_fifo_errors = b->rx_fifo_errors;
 898         a->rx_missed_errors = b->rx_missed_errors;
 899
 900         a->tx_aborted_errors = b->tx_aborted_errors;
 901         a->tx_carrier_errors = b->tx_carrier_errors;
 902         a->tx_fifo_errors = b->tx_fifo_errors;
 903         a->tx_heartbeat_errors = b->tx_heartbeat_errors;
 904         a->tx_window_errors = b->tx_window_errors;
 905
 906         a->rx_compressed = b->rx_compressed;
 907         a->tx_compressed = b->tx_compressed;
 908
 909         a->rx_nohandler = b->rx_nohandler;
 910 }
 911
 912 /* All VF info */
 913 static inline int rtnl_vfinfo_size(const struct net_device *dev,
 914                                    u32 ext_filter_mask)
 915 {
 916         if (dev->dev.parent && (ext_filter_mask & RTEXT_FILTER_VF)) {
 917                 int num_vfs = dev_num_vf(dev->dev.parent);
 918                 size_t size = nla_total_size(0);
 919                 size += num_vfs *
 920                         (nla_total_size(0) +
 921                          nla_total_size(sizeof(struct ifla_vf_mac)) +
 922                          nla_total_size(sizeof(struct ifla_vf_broadcast)) +
 923                          nla_total_size(sizeof(struct ifla_vf_vlan)) +
 924                          nla_total_size(0) + /* nest IFLA_VF_VLAN_LIST */
 925                          nla_total_size(MAX_VLAN_LIST_LEN *
 926                                         sizeof(struct ifla_vf_vlan_info)) +
 927                          nla_total_size(sizeof(struct ifla_vf_spoofchk)) +
 928                          nla_total_size(sizeof(struct ifla_vf_tx_rate)) +
 929                          nla_total_size(sizeof(struct ifla_vf_rate)) +
 930                          nla_total_size(sizeof(struct ifla_vf_link_state)) +
 931                          nla_total_size(sizeof(struct ifla_vf_rss_query_en)) +
 932                          nla_total_size(0) + /* nest IFLA_VF_STATS */
 933                          /* IFLA_VF_STATS_RX_PACKETS */
 934                          nla_total_size_64bit(sizeof(__u64)) +
 935                          /* IFLA_VF_STATS_TX_PACKETS */
 936                          nla_total_size_64bit(sizeof(__u64)) +
 937                          /* IFLA_VF_STATS_RX_BYTES */
 938                          nla_total_size_64bit(sizeof(__u64)) +
 939                          /* IFLA_VF_STATS_TX_BYTES */
 940                          nla_total_size_64bit(sizeof(__u64)) +
 941                          /* IFLA_VF_STATS_BROADCAST */
 942                          nla_total_size_64bit(sizeof(__u64)) +
 943                          /* IFLA_VF_STATS_MULTICAST */
 944                          nla_total_size_64bit(sizeof(__u64)) +
 945                          /* IFLA_VF_STATS_RX_DROPPED */
 946                          nla_total_size_64bit(sizeof(__u64)) +
 947                          /* IFLA_VF_STATS_TX_DROPPED */
 948                          nla_total_size_64bit(sizeof(__u64)) +
 949                          nla_total_size(sizeof(struct ifla_vf_trust)));
 950                 return size;
 951         } else
 952                 return 0;
 953 }
 954
 955 static size_t rtnl_port_size(const struct net_device *dev,
 956                              u32 ext_filter_mask)
 957 {
 958         size_t port_size = nla_total_size(4)            /* PORT_VF */
 959                 + nla_total_size(PORT_PROFILE_MAX)      /* PORT_PROFILE */
 960                 + nla_total_size(PORT_UUID_MAX)         /* PORT_INSTANCE_UUID */
 961                 + nla_total_size(PORT_UUID_MAX)         /* PORT_HOST_UUID */
 962                 + nla_total_size(1)                     /* PROT_VDP_REQUEST */
 963                 + nla_total_size(2);                    /* PORT_VDP_RESPONSE */
 964         size_t vf_ports_size = nla_total_size(sizeof(struct nlattr));
 965         size_t vf_port_size = nla_total_size(sizeof(struct nlattr))
 966                 + port_size;
 967         size_t port_self_size = nla_total_size(sizeof(struct nlattr))
 968                 + port_size;
 969
 970         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
 971             !(ext_filter_mask & RTEXT_FILTER_VF))
 972                 return 0;
 973         if (dev_num_vf(dev->dev.parent))
 974                 return port_self_size + vf_ports_size +
 975                         vf_port_size * dev_num_vf(dev->dev.parent);
 976         else
 977                 return port_self_size;
 978 }
 979
 980 static size_t rtnl_xdp_size(void)
 981 {
 982         size_t xdp_size = nla_total_size(0) +   /* nest IFLA_XDP */
 983                           nla_total_size(1) +   /* XDP_ATTACHED */
 984                           nla_total_size(4) +   /* XDP_PROG_ID (or 1st mode) */
 985                           nla_total_size(4);    /* XDP_<mode>_PROG_ID */
 986
 987         return xdp_size;
 988 }
 989
 990 static size_t rtnl_prop_list_size(const struct net_device *dev)
 991 {
 992         struct netdev_name_node *name_node;
 993         size_t size;
 994
 995         if (list_empty(&dev->name_node->list))
 996                 return 0;
 997         size = nla_total_size(0);
 998         list_for_each_entry(name_node, &dev->name_node->list, list)
 999                 size += nla_total_size(ALTIFNAMSIZ);
1000         return size;
1001 }
1002
1003 static size_t rtnl_proto_down_size(const struct net_device *dev)
1004 {
1005         size_t size = nla_total_size(1);
1006
1007         if (dev->proto_down_reason)
1008                 size += nla_total_size(0) + nla_total_size(4);
1009
1010         return size;
1011 }
1012
1013 static noinline size_t if_nlmsg_size(const struct net_device *dev,
1014                                      u32 ext_filter_mask)
1015 {
1016         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
1017                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
1018                + nla_total_size(IFALIASZ) /* IFLA_IFALIAS */
1019                + nla_total_size(IFNAMSIZ) /* IFLA_QDISC */
1020                + nla_total_size_64bit(sizeof(struct rtnl_link_ifmap))
1021                + nla_total_size(sizeof(struct rtnl_link_stats))
1022                + nla_total_size_64bit(sizeof(struct rtnl_link_stats64))
1023                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
1024                + nla_total_size(MAX_ADDR_LEN) /* IFLA_BROADCAST */
1025                + nla_total_size(4) /* IFLA_TXQLEN */
1026                + nla_total_size(4) /* IFLA_WEIGHT */
1027                + nla_total_size(4) /* IFLA_MTU */
1028                + nla_total_size(4) /* IFLA_LINK */
1029                + nla_total_size(4) /* IFLA_MASTER */
1030                + nla_total_size(1) /* IFLA_CARRIER */
1031                + nla_total_size(4) /* IFLA_PROMISCUITY */
1032                + nla_total_size(4) /* IFLA_NUM_TX_QUEUES */
1033                + nla_total_size(4) /* IFLA_NUM_RX_QUEUES */
1034                + nla_total_size(4) /* IFLA_GSO_MAX_SEGS */
1035                + nla_total_size(4) /* IFLA_GSO_MAX_SIZE */
1036                + nla_total_size(1) /* IFLA_OPERSTATE */
1037                + nla_total_size(1) /* IFLA_LINKMODE */
1038                + nla_total_size(4) /* IFLA_CARRIER_CHANGES */
1039                + nla_total_size(4) /* IFLA_LINK_NETNSID */
1040                + nla_total_size(4) /* IFLA_GROUP */
1041                + nla_total_size(ext_filter_mask
1042                                 & RTEXT_FILTER_VF ? 4 : 0) /* IFLA_NUM_VF */
1043                + rtnl_vfinfo_size(dev, ext_filter_mask) /* IFLA_VFINFO_LIST */
1044                + rtnl_port_size(dev, ext_filter_mask) /* IFLA_VF_PORTS + IFLA_PORT_SELF */
1045                + rtnl_link_get_size(dev) /* IFLA_LINKINFO */
1046                + rtnl_link_get_af_size(dev, ext_filter_mask) /* IFLA_AF_SPEC */
1047                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_PORT_ID */
1048                + nla_total_size(MAX_PHYS_ITEM_ID_LEN) /* IFLA_PHYS_SWITCH_ID */
1049                + nla_total_size(IFNAMSIZ) /* IFLA_PHYS_PORT_NAME */
1050                + rtnl_xdp_size() /* IFLA_XDP */
1051                + nla_total_size(4)  /* IFLA_EVENT */
1052                + nla_total_size(4)  /* IFLA_NEW_NETNSID */
1053                + nla_total_size(4)  /* IFLA_NEW_IFINDEX */
1054                + rtnl_proto_down_size(dev)  /* proto down */
1055                + nla_total_size(4)  /* IFLA_TARGET_NETNSID */
1056                + nla_total_size(4)  /* IFLA_CARRIER_UP_COUNT */
1057                + nla_total_size(4)  /* IFLA_CARRIER_DOWN_COUNT */
1058                + nla_total_size(4)  /* IFLA_MIN_MTU */
1059                + nla_total_size(4)  /* IFLA_MAX_MTU */
1060                + rtnl_prop_list_size(dev)
1061                + nla_total_size(MAX_ADDR_LEN) /* IFLA_PERM_ADDRESS */
1062                + 0;
1063 }
1064
1065 static int rtnl_vf_ports_fill(struct sk_buff *skb, struct net_device *dev)
1066 {
1067         struct nlattr *vf_ports;
1068         struct nlattr *vf_port;
1069         int vf;
1070         int err;
1071
1072         vf_ports = nla_nest_start_noflag(skb, IFLA_VF_PORTS);
1073         if (!vf_ports)
1074                 return -EMSGSIZE;
1075
1076         for (vf = 0; vf < dev_num_vf(dev->dev.parent); vf++) {
1077                 vf_port = nla_nest_start_noflag(skb, IFLA_VF_PORT);
1078                 if (!vf_port)
1079                         goto nla_put_failure;
1080                 if (nla_put_u32(skb, IFLA_PORT_VF, vf))
1081                         goto nla_put_failure;
1082                 err = dev->netdev_ops->ndo_get_vf_port(dev, vf, skb);
1083                 if (err == -EMSGSIZE)
1084                         goto nla_put_failure;
1085                 if (err) {
1086                         nla_nest_cancel(skb, vf_port);
1087                         continue;
1088                 }
1089                 nla_nest_end(skb, vf_port);
1090         }
1091
1092         nla_nest_end(skb, vf_ports);
1093
1094         return 0;
1095
1096 nla_put_failure:
1097         nla_nest_cancel(skb, vf_ports);
1098         return -EMSGSIZE;
1099 }
1100
1101 static int rtnl_port_self_fill(struct sk_buff *skb, struct net_device *dev)
1102 {
1103         struct nlattr *port_self;
1104         int err;
1105
1106         port_self = nla_nest_start_noflag(skb, IFLA_PORT_SELF);
1107         if (!port_self)
1108                 return -EMSGSIZE;
1109
1110         err = dev->netdev_ops->ndo_get_vf_port(dev, PORT_SELF_VF, skb);
1111         if (err) {
1112                 nla_nest_cancel(skb, port_self);
1113                 return (err == -EMSGSIZE) ? err : 0;
1114         }
1115
1116         nla_nest_end(skb, port_self);
1117
1118         return 0;
1119 }
1120
1121 static int rtnl_port_fill(struct sk_buff *skb, struct net_device *dev,
1122                           u32 ext_filter_mask)
1123 {
1124         int err;
1125
1126         if (!dev->netdev_ops->ndo_get_vf_port || !dev->dev.parent ||
1127             !(ext_filter_mask & RTEXT_FILTER_VF))
1128                 return 0;
1129
1130         err = rtnl_port_self_fill(skb, dev);
1131         if (err)
1132                 return err;
1133
1134         if (dev_num_vf(dev->dev.parent)) {
1135                 err = rtnl_vf_ports_fill(skb, dev);
1136                 if (err)
1137                         return err;
1138         }
1139
1140         return 0;
1141 }
1142
1143 static int rtnl_phys_port_id_fill(struct sk_buff *skb, struct net_device *dev)
1144 {
1145         int err;
1146         struct netdev_phys_item_id ppid;
1147
1148         err = dev_get_phys_port_id(dev, &ppid);
1149         if (err) {
1150                 if (err == -EOPNOTSUPP)
1151                         return 0;
1152                 return err;
1153         }
1154
1155         if (nla_put(skb, IFLA_PHYS_PORT_ID, ppid.id_len, ppid.id))
1156                 return -EMSGSIZE;
1157
1158         return 0;
1159 }
1160
1161 static int rtnl_phys_port_name_fill(struct sk_buff *skb, struct net_device *dev)
1162 {
1163         char name[IFNAMSIZ];
1164         int err;
1165
1166         err = dev_get_phys_port_name(dev, name, sizeof(name));
1167         if (err) {
1168                 if (err == -EOPNOTSUPP)
1169                         return 0;
1170                 return err;
1171         }
1172
1173         if (nla_put_string(skb, IFLA_PHYS_PORT_NAME, name))
1174                 return -EMSGSIZE;
1175
1176         return 0;
1177 }
1178
1179 static int rtnl_phys_switch_id_fill(struct sk_buff *skb, struct net_device *dev)
1180 {
1181         struct netdev_phys_item_id ppid = { };
1182         int err;
1183
1184         err = dev_get_port_parent_id(dev, &ppid, false);
1185         if (err) {
1186                 if (err == -EOPNOTSUPP)
1187                         return 0;
1188                 return err;
1189         }
1190
1191         if (nla_put(skb, IFLA_PHYS_SWITCH_ID, ppid.id_len, ppid.id))
1192                 return -EMSGSIZE;
1193
1194         return 0;
1195 }
1196
1197 static noinline_for_stack int rtnl_fill_stats(struct sk_buff *skb,
1198                                               struct net_device *dev)
1199 {
1200         struct rtnl_link_stats64 *sp;
1201         struct nlattr *attr;
1202
1203         attr = nla_reserve_64bit(skb, IFLA_STATS64,
1204                                  sizeof(struct rtnl_link_stats64), IFLA_PAD);
1205         if (!attr)
1206                 return -EMSGSIZE;
1207
1208         sp = nla_data(attr);
1209         dev_get_stats(dev, sp);
1210
1211         attr = nla_reserve(skb, IFLA_STATS,
1212                            sizeof(struct rtnl_link_stats));
1213         if (!attr)
1214                 return -EMSGSIZE;
1215
1216         copy_rtnl_link_stats(nla_data(attr), sp);
1217
1218         return 0;
1219 }
1220
1221 static noinline_for_stack int rtnl_fill_vfinfo(struct sk_buff *skb,
1222                                                struct net_device *dev,
1223                                                int vfs_num,
1224                                                struct nlattr *vfinfo)
1225 {
1226         struct ifla_vf_rss_query_en vf_rss_query_en;
1227         struct nlattr *vf, *vfstats, *vfvlanlist;
1228         struct ifla_vf_link_state vf_linkstate;
1229         struct ifla_vf_vlan_info vf_vlan_info;
1230         struct ifla_vf_spoofchk vf_spoofchk;
1231         struct ifla_vf_tx_rate vf_tx_rate;
1232         struct ifla_vf_stats vf_stats;
1233         struct ifla_vf_trust vf_trust;
1234         struct ifla_vf_vlan vf_vlan;
1235         struct ifla_vf_rate vf_rate;
1236         struct ifla_vf_mac vf_mac;
1237         struct ifla_vf_broadcast vf_broadcast;
1238         struct ifla_vf_info ivi;
1239         struct ifla_vf_guid node_guid;
1240         struct ifla_vf_guid port_guid;
1241
1242         memset(&ivi, 0, sizeof(ivi));
1243
1244         /* Not all SR-IOV capable drivers support the
1245          * spoofcheck and "RSS query enable" query.  Preset to
1246          * -1 so the user space tool can detect that the driver
1247          * didn't report anything.
1248          */
1249         ivi.spoofchk = -1;
1250         ivi.rss_query_en = -1;
1251         ivi.trusted = -1;
1252         /* The default value for VF link state is "auto"
1253          * IFLA_VF_LINK_STATE_AUTO which equals zero
1254          */
1255         ivi.linkstate = 0;
1256         /* VLAN Protocol by default is 802.1Q */
1257         ivi.vlan_proto = htons(ETH_P_8021Q);
1258         if (dev->netdev_ops->ndo_get_vf_config(dev, vfs_num, &ivi))
1259                 return 0;
1260
1261         memset(&vf_vlan_info, 0, sizeof(vf_vlan_info));
1262         memset(&node_guid, 0, sizeof(node_guid));
1263         memset(&port_guid, 0, sizeof(port_guid));
1264
1265         vf_mac.vf =
1266                 vf_vlan.vf =
1267                 vf_vlan_info.vf =
1268                 vf_rate.vf =
1269                 vf_tx_rate.vf =
1270                 vf_spoofchk.vf =
1271                 vf_linkstate.vf =
1272                 vf_rss_query_en.vf =
1273                 vf_trust.vf =
1274                 node_guid.vf =
1275                 port_guid.vf = ivi.vf;
1276
1277         memcpy(vf_mac.mac, ivi.mac, sizeof(ivi.mac));
1278         memcpy(vf_broadcast.broadcast, dev->broadcast, dev->addr_len);
1279         vf_vlan.vlan = ivi.vlan;
1280         vf_vlan.qos = ivi.qos;
1281         vf_vlan_info.vlan = ivi.vlan;
1282         vf_vlan_info.qos = ivi.qos;
1283         vf_vlan_info.vlan_proto = ivi.vlan_proto;
1284         vf_tx_rate.rate = ivi.max_tx_rate;
1285         vf_rate.min_tx_rate = ivi.min_tx_rate;
1286         vf_rate.max_tx_rate = ivi.max_tx_rate;
1287         vf_spoofchk.setting = ivi.spoofchk;
1288         vf_linkstate.link_state = ivi.linkstate;
1289         vf_rss_query_en.setting = ivi.rss_query_en;
1290         vf_trust.setting = ivi.trusted;
1291         vf = nla_nest_start_noflag(skb, IFLA_VF_INFO);
1292         if (!vf)
1293                 goto nla_put_vfinfo_failure;
1294         if (nla_put(skb, IFLA_VF_MAC, sizeof(vf_mac), &vf_mac) ||
1295             nla_put(skb, IFLA_VF_BROADCAST, sizeof(vf_broadcast), &vf_broadcast) ||
1296             nla_put(skb, IFLA_VF_VLAN, sizeof(vf_vlan), &vf_vlan) ||
1297             nla_put(skb, IFLA_VF_RATE, sizeof(vf_rate),
1298                     &vf_rate) ||
1299             nla_put(skb, IFLA_VF_TX_RATE, sizeof(vf_tx_rate),
1300                     &vf_tx_rate) ||
1301             nla_put(skb, IFLA_VF_SPOOFCHK, sizeof(vf_spoofchk),
1302                     &vf_spoofchk) ||
1303             nla_put(skb, IFLA_VF_LINK_STATE, sizeof(vf_linkstate),
1304                     &vf_linkstate) ||
1305             nla_put(skb, IFLA_VF_RSS_QUERY_EN,
1306                     sizeof(vf_rss_query_en),
1307                     &vf_rss_query_en) ||
1308             nla_put(skb, IFLA_VF_TRUST,
1309                     sizeof(vf_trust), &vf_trust))
1310                 goto nla_put_vf_failure;
1311
1312         if (dev->netdev_ops->ndo_get_vf_guid &&
1313             !dev->netdev_ops->ndo_get_vf_guid(dev, vfs_num, &node_guid,
1314                                               &port_guid)) {
1315                 if (nla_put(skb, IFLA_VF_IB_NODE_GUID, sizeof(node_guid),
1316                             &node_guid) ||
1317                     nla_put(skb, IFLA_VF_IB_PORT_GUID, sizeof(port_guid),
1318                             &port_guid))
1319                         goto nla_put_vf_failure;
1320         }
1321         vfvlanlist = nla_nest_start_noflag(skb, IFLA_VF_VLAN_LIST);
1322         if (!vfvlanlist)
1323                 goto nla_put_vf_failure;
1324         if (nla_put(skb, IFLA_VF_VLAN_INFO, sizeof(vf_vlan_info),
1325                     &vf_vlan_info)) {
1326                 nla_nest_cancel(skb, vfvlanlist);
1327                 goto nla_put_vf_failure;
1328         }
1329         nla_nest_end(skb, vfvlanlist);
1330         memset(&vf_stats, 0, sizeof(vf_stats));
1331         if (dev->netdev_ops->ndo_get_vf_stats)
1332                 dev->netdev_ops->ndo_get_vf_stats(dev, vfs_num,
1333                                                 &vf_stats);
1334         vfstats = nla_nest_start_noflag(skb, IFLA_VF_STATS);
1335         if (!vfstats)
1336                 goto nla_put_vf_failure;
1337         if (nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_PACKETS,
1338                               vf_stats.rx_packets, IFLA_VF_STATS_PAD) ||
1339             nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_PACKETS,
1340                               vf_stats.tx_packets, IFLA_VF_STATS_PAD) ||
1341             nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_BYTES,
1342                               vf_stats.rx_bytes, IFLA_VF_STATS_PAD) ||
1343             nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_BYTES,
1344                               vf_stats.tx_bytes, IFLA_VF_STATS_PAD) ||
1345             nla_put_u64_64bit(skb, IFLA_VF_STATS_BROADCAST,
1346                               vf_stats.broadcast, IFLA_VF_STATS_PAD) ||
1347             nla_put_u64_64bit(skb, IFLA_VF_STATS_MULTICAST,
1348                               vf_stats.multicast, IFLA_VF_STATS_PAD) ||
1349             nla_put_u64_64bit(skb, IFLA_VF_STATS_RX_DROPPED,
1350                               vf_stats.rx_dropped, IFLA_VF_STATS_PAD) ||
1351             nla_put_u64_64bit(skb, IFLA_VF_STATS_TX_DROPPED,
1352                               vf_stats.tx_dropped, IFLA_VF_STATS_PAD)) {
1353                 nla_nest_cancel(skb, vfstats);
1354                 goto nla_put_vf_failure;
1355         }
1356         nla_nest_end(skb, vfstats);
1357         nla_nest_end(skb, vf);
1358         return 0;
1359
1360 nla_put_vf_failure:
1361         nla_nest_cancel(skb, vf);
1362 nla_put_vfinfo_failure:
1363         nla_nest_cancel(skb, vfinfo);
1364         return -EMSGSIZE;
1365 }
1366
1367 static noinline_for_stack int rtnl_fill_vf(struct sk_buff *skb,
1368                                            struct net_device *dev,
1369                                            u32 ext_filter_mask)
1370 {
1371         struct nlattr *vfinfo;
1372         int i, num_vfs;
1373
1374         if (!dev->dev.parent || ((ext_filter_mask & RTEXT_FILTER_VF) == 0))
1375                 return 0;
1376
1377         num_vfs = dev_num_vf(dev->dev.parent);
1378         if (nla_put_u32(skb, IFLA_NUM_VF, num_vfs))
1379                 return -EMSGSIZE;
1380
1381         if (!dev->netdev_ops->ndo_get_vf_config)
1382                 return 0;
1383
1384         vfinfo = nla_nest_start_noflag(skb, IFLA_VFINFO_LIST);
1385         if (!vfinfo)
1386                 return -EMSGSIZE;
1387
1388         for (i = 0; i < num_vfs; i++) {
1389                 if (rtnl_fill_vfinfo(skb, dev, i, vfinfo))
1390                         return -EMSGSIZE;
1391         }
1392
1393         nla_nest_end(skb, vfinfo);
1394         return 0;
1395 }
1396
1397 static int rtnl_fill_link_ifmap(struct sk_buff *skb, struct net_device *dev)
1398 {
1399         struct rtnl_link_ifmap map;
1400
1401         memset(&map, 0, sizeof(map));
1402         map.mem_start   = dev->mem_start;
1403         map.mem_end     = dev->mem_end;
1404         map.base_addr   = dev->base_addr;
1405         map.irq         = dev->irq;
1406         map.dma         = dev->dma;
1407         map.port        = dev->if_port;
1408
1409         if (nla_put_64bit(skb, IFLA_MAP, sizeof(map), &map, IFLA_PAD))
1410                 return -EMSGSIZE;
1411
1412         return 0;
1413 }
1414
1415 static u32 rtnl_xdp_prog_skb(struct net_device *dev)
1416 {
1417         const struct bpf_prog *generic_xdp_prog;
1418
1419         ASSERT_RTNL();
1420
1421         generic_xdp_prog = rtnl_dereference(dev->xdp_prog);
1422         if (!generic_xdp_prog)
1423                 return 0;
1424         return generic_xdp_prog->aux->id;
1425 }
1426
1427 static u32 rtnl_xdp_prog_drv(struct net_device *dev)
1428 {
1429         return dev_xdp_prog_id(dev, XDP_MODE_DRV);
1430 }
1431
1432 static u32 rtnl_xdp_prog_hw(struct net_device *dev)
1433 {
1434         return dev_xdp_prog_id(dev, XDP_MODE_HW);
1435 }
1436
1437 static int rtnl_xdp_report_one(struct sk_buff *skb, struct net_device *dev,
1438                                u32 *prog_id, u8 *mode, u8 tgt_mode, u32 attr,
1439                                u32 (*get_prog_id)(struct net_device *dev))
1440 {
1441         u32 curr_id;
1442         int err;
1443
1444         curr_id = get_prog_id(dev);
1445         if (!curr_id)
1446                 return 0;
1447
1448         *prog_id = curr_id;
1449         err = nla_put_u32(skb, attr, curr_id);
1450         if (err)
1451                 return err;
1452
1453         if (*mode != XDP_ATTACHED_NONE)
1454                 *mode = XDP_ATTACHED_MULTI;
1455         else
1456                 *mode = tgt_mode;
1457
1458         return 0;
1459 }
1460
1461 static int rtnl_xdp_fill(struct sk_buff *skb, struct net_device *dev)
1462 {
1463         struct nlattr *xdp;
1464         u32 prog_id;
1465         int err;
1466         u8 mode;
1467
1468         xdp = nla_nest_start_noflag(skb, IFLA_XDP);
1469         if (!xdp)
1470                 return -EMSGSIZE;
1471
1472         prog_id = 0;
1473         mode = XDP_ATTACHED_NONE;
1474         err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_SKB,
1475                                   IFLA_XDP_SKB_PROG_ID, rtnl_xdp_prog_skb);
1476         if (err)
1477                 goto err_cancel;
1478         err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_DRV,
1479                                   IFLA_XDP_DRV_PROG_ID, rtnl_xdp_prog_drv);
1480         if (err)
1481                 goto err_cancel;
1482         err = rtnl_xdp_report_one(skb, dev, &prog_id, &mode, XDP_ATTACHED_HW,
1483                                   IFLA_XDP_HW_PROG_ID, rtnl_xdp_prog_hw);
1484         if (err)
1485                 goto err_cancel;
1486
1487         err = nla_put_u8(skb, IFLA_XDP_ATTACHED, mode);
1488         if (err)
1489                 goto err_cancel;
1490
1491         if (prog_id && mode != XDP_ATTACHED_MULTI) {
1492                 err = nla_put_u32(skb, IFLA_XDP_PROG_ID, prog_id);
1493                 if (err)
1494                         goto err_cancel;
1495         }
1496
1497         nla_nest_end(skb, xdp);
1498         return 0;
1499
1500 err_cancel:
1501         nla_nest_cancel(skb, xdp);
1502         return err;
1503 }
1504
1505 static u32 rtnl_get_event(unsigned long event)
1506 {
1507         u32 rtnl_event_type = IFLA_EVENT_NONE;
1508
1509         switch (event) {
1510         case NETDEV_REBOOT:
1511                 rtnl_event_type = IFLA_EVENT_REBOOT;
1512                 break;
1513         case NETDEV_FEAT_CHANGE:
1514                 rtnl_event_type = IFLA_EVENT_FEATURES;
1515                 break;
1516         case NETDEV_BONDING_FAILOVER:
1517                 rtnl_event_type = IFLA_EVENT_BONDING_FAILOVER;
1518                 break;
1519         case NETDEV_NOTIFY_PEERS:
1520                 rtnl_event_type = IFLA_EVENT_NOTIFY_PEERS;
1521                 break;
1522         case NETDEV_RESEND_IGMP:
1523                 rtnl_event_type = IFLA_EVENT_IGMP_RESEND;
1524                 break;
1525         case NETDEV_CHANGEINFODATA:
1526                 rtnl_event_type = IFLA_EVENT_BONDING_OPTIONS;
1527                 break;
1528         default:
1529                 break;
1530         }
1531
1532         return rtnl_event_type;
1533 }
1534
1535 static int put_master_ifindex(struct sk_buff *skb, struct net_device *dev)
1536 {
1537         const struct net_device *upper_dev;
1538         int ret = 0;
1539
1540         rcu_read_lock();
1541
1542         upper_dev = netdev_master_upper_dev_get_rcu(dev);
1543         if (upper_dev)
1544                 ret = nla_put_u32(skb, IFLA_MASTER, upper_dev->ifindex);
1545
1546         rcu_read_unlock();
1547         return ret;
1548 }
1549
1550 static int nla_put_iflink(struct sk_buff *skb, const struct net_device *dev,
1551                           bool force)
1552 {
1553         int ifindex = dev_get_iflink(dev);
1554
1555         if (force || dev->ifindex != ifindex)
1556                 return nla_put_u32(skb, IFLA_LINK, ifindex);
1557
1558         return 0;
1559 }
1560
1561 static noinline_for_stack int nla_put_ifalias(struct sk_buff *skb,
1562                                               struct net_device *dev)
1563 {
1564         char buf[IFALIASZ];
1565         int ret;
1566
1567         ret = dev_get_alias(dev, buf, sizeof(buf));
1568         return ret > 0 ? nla_put_string(skb, IFLA_IFALIAS, buf) : 0;
1569 }
1570
1571 static int rtnl_fill_link_netnsid(struct sk_buff *skb,
1572                                   const struct net_device *dev,
1573                                   struct net *src_net, gfp_t gfp)
1574 {
1575         bool put_iflink = false;
1576
1577         if (dev->rtnl_link_ops && dev->rtnl_link_ops->get_link_net) {
1578                 struct net *link_net = dev->rtnl_link_ops->get_link_net(dev);
1579
1580                 if (!net_eq(dev_net(dev), link_net)) {
1581                         int id = peernet2id_alloc(src_net, link_net, gfp);
1582
1583                         if (nla_put_s32(skb, IFLA_LINK_NETNSID, id))
1584                                 return -EMSGSIZE;
1585
1586                         put_iflink = true;
1587                 }
1588         }
1589
1590         return nla_put_iflink(skb, dev, put_iflink);
1591 }
1592
1593 static int rtnl_fill_link_af(struct sk_buff *skb,
1594                              const struct net_device *dev,
1595                              u32 ext_filter_mask)
1596 {
1597         const struct rtnl_af_ops *af_ops;
1598         struct nlattr *af_spec;
1599
1600         af_spec = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
1601         if (!af_spec)
1602                 return -EMSGSIZE;
1603
1604         list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
1605                 struct nlattr *af;
1606                 int err;
1607
1608                 if (!af_ops->fill_link_af)
1609                         continue;
1610
1611                 af = nla_nest_start_noflag(skb, af_ops->family);
1612                 if (!af)
1613                         return -EMSGSIZE;
1614
1615                 err = af_ops->fill_link_af(skb, dev, ext_filter_mask);
1616                 /*
1617                  * Caller may return ENODATA to indicate that there
1618                  * was no data to be dumped. This is not an error, it
1619                  * means we should trim the attribute header and
1620                  * continue.
1621                  */
1622                 if (err == -ENODATA)
1623                         nla_nest_cancel(skb, af);
1624                 else if (err < 0)
1625                         return -EMSGSIZE;
1626
1627                 nla_nest_end(skb, af);
1628         }
1629
1630         nla_nest_end(skb, af_spec);
1631         return 0;
1632 }
1633
1634 static int rtnl_fill_alt_ifnames(struct sk_buff *skb,
1635                                  const struct net_device *dev)
1636 {
1637         struct netdev_name_node *name_node;
1638         int count = 0;
1639
1640         list_for_each_entry(name_node, &dev->name_node->list, list) {
1641                 if (nla_put_string(skb, IFLA_ALT_IFNAME, name_node->name))
1642                         return -EMSGSIZE;
1643                 count++;
1644         }
1645         return count;
1646 }
1647
1648 static int rtnl_fill_prop_list(struct sk_buff *skb,
1649                                const struct net_device *dev)
1650 {
1651         struct nlattr *prop_list;
1652         int ret;
1653
1654         prop_list = nla_nest_start(skb, IFLA_PROP_LIST);
1655         if (!prop_list)
1656                 return -EMSGSIZE;
1657
1658         ret = rtnl_fill_alt_ifnames(skb, dev);
1659         if (ret <= 0)
1660                 goto nest_cancel;
1661
1662         nla_nest_end(skb, prop_list);
1663         return 0;
1664
1665 nest_cancel:
1666         nla_nest_cancel(skb, prop_list);
1667         return ret;
1668 }
1669
1670 static int rtnl_fill_proto_down(struct sk_buff *skb,
1671                                 const struct net_device *dev)
1672 {
1673         struct nlattr *pr;
1674         u32 preason;
1675
1676         if (nla_put_u8(skb, IFLA_PROTO_DOWN, dev->proto_down))
1677                 goto nla_put_failure;
1678
1679         preason = dev->proto_down_reason;
1680         if (!preason)
1681                 return 0;
1682
1683         pr = nla_nest_start(skb, IFLA_PROTO_DOWN_REASON);
1684         if (!pr)
1685                 return -EMSGSIZE;
1686
1687         if (nla_put_u32(skb, IFLA_PROTO_DOWN_REASON_VALUE, preason)) {
1688                 nla_nest_cancel(skb, pr);
1689                 goto nla_put_failure;
1690         }
1691
1692         nla_nest_end(skb, pr);
1693         return 0;
1694
1695 nla_put_failure:
1696         return -EMSGSIZE;
1697 }
1698
1699 static int rtnl_fill_ifinfo(struct sk_buff *skb,
1700                             struct net_device *dev, struct net *src_net,
1701                             int type, u32 pid, u32 seq, u32 change,
1702                             unsigned int flags, u32 ext_filter_mask,
1703                             u32 event, int *new_nsid, int new_ifindex,
1704                             int tgt_netnsid, gfp_t gfp)
1705 {
1706         struct ifinfomsg *ifm;
1707         struct nlmsghdr *nlh;
1708         struct Qdisc *qdisc;
1709
1710         ASSERT_RTNL();
1711         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifm), flags);
1712         if (nlh == NULL)
1713                 return -EMSGSIZE;
1714
1715         ifm = nlmsg_data(nlh);
1716         ifm->ifi_family = AF_UNSPEC;
1717         ifm->__ifi_pad = 0;
1718         ifm->ifi_type = dev->type;
1719         ifm->ifi_index = dev->ifindex;
1720         ifm->ifi_flags = dev_get_flags(dev);
1721         ifm->ifi_change = change;
1722
1723         if (tgt_netnsid >= 0 && nla_put_s32(skb, IFLA_TARGET_NETNSID, tgt_netnsid))
1724                 goto nla_put_failure;
1725
1726         qdisc = rtnl_dereference(dev->qdisc);
1727         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
1728             nla_put_u32(skb, IFLA_TXQLEN, dev->tx_queue_len) ||
1729             nla_put_u8(skb, IFLA_OPERSTATE,
1730                        netif_running(dev) ? dev->operstate : IF_OPER_DOWN) ||
1731             nla_put_u8(skb, IFLA_LINKMODE, dev->link_mode) ||
1732             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
1733             nla_put_u32(skb, IFLA_MIN_MTU, dev->min_mtu) ||
1734             nla_put_u32(skb, IFLA_MAX_MTU, dev->max_mtu) ||
1735             nla_put_u32(skb, IFLA_GROUP, dev->group) ||
1736             nla_put_u32(skb, IFLA_PROMISCUITY, dev->promiscuity) ||
1737             nla_put_u32(skb, IFLA_NUM_TX_QUEUES, dev->num_tx_queues) ||
1738             nla_put_u32(skb, IFLA_GSO_MAX_SEGS, dev->gso_max_segs) ||
1739             nla_put_u32(skb, IFLA_GSO_MAX_SIZE, dev->gso_max_size) ||
1740 #ifdef CONFIG_RPS
1741             nla_put_u32(skb, IFLA_NUM_RX_QUEUES, dev->num_rx_queues) ||
1742 #endif
1743             put_master_ifindex(skb, dev) ||
1744             nla_put_u8(skb, IFLA_CARRIER, netif_carrier_ok(dev)) ||
1745             (qdisc &&
1746              nla_put_string(skb, IFLA_QDISC, qdisc->ops->id)) ||
1747             nla_put_ifalias(skb, dev) ||
1748             nla_put_u32(skb, IFLA_CARRIER_CHANGES,
1749                         atomic_read(&dev->carrier_up_count) +
1750                         atomic_read(&dev->carrier_down_count)) ||
1751             nla_put_u32(skb, IFLA_CARRIER_UP_COUNT,
1752                         atomic_read(&dev->carrier_up_count)) ||
1753             nla_put_u32(skb, IFLA_CARRIER_DOWN_COUNT,
1754                         atomic_read(&dev->carrier_down_count)))
1755                 goto nla_put_failure;
1756
1757         if (rtnl_fill_proto_down(skb, dev))
1758                 goto nla_put_failure;
1759
1760         if (event != IFLA_EVENT_NONE) {
1761                 if (nla_put_u32(skb, IFLA_EVENT, event))
1762                         goto nla_put_failure;
1763         }
1764
1765         if (rtnl_fill_link_ifmap(skb, dev))
1766                 goto nla_put_failure;
1767
1768         if (dev->addr_len) {
1769                 if (nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr) ||
1770                     nla_put(skb, IFLA_BROADCAST, dev->addr_len, dev->broadcast))
1771                         goto nla_put_failure;
1772         }
1773
1774         if (rtnl_phys_port_id_fill(skb, dev))
1775                 goto nla_put_failure;
1776
1777         if (rtnl_phys_port_name_fill(skb, dev))
1778                 goto nla_put_failure;
1779
1780         if (rtnl_phys_switch_id_fill(skb, dev))
1781                 goto nla_put_failure;
1782
1783         if (rtnl_fill_stats(skb, dev))
1784                 goto nla_put_failure;
1785
1786         if (rtnl_fill_vf(skb, dev, ext_filter_mask))
1787                 goto nla_put_failure;
1788
1789         if (rtnl_port_fill(skb, dev, ext_filter_mask))
1790                 goto nla_put_failure;
1791
1792         if (rtnl_xdp_fill(skb, dev))
1793                 goto nla_put_failure;
1794
1795         if (dev->rtnl_link_ops || rtnl_have_link_slave_info(dev)) {
1796                 if (rtnl_link_fill(skb, dev) < 0)
1797                         goto nla_put_failure;
1798         }
1799
1800         if (rtnl_fill_link_netnsid(skb, dev, src_net, gfp))
1801                 goto nla_put_failure;
1802
1803         if (new_nsid &&
1804             nla_put_s32(skb, IFLA_NEW_NETNSID, *new_nsid) < 0)
1805                 goto nla_put_failure;
1806         if (new_ifindex &&
1807             nla_put_s32(skb, IFLA_NEW_IFINDEX, new_ifindex) < 0)
1808                 goto nla_put_failure;
1809
1810         if (memchr_inv(dev->perm_addr, '\0', dev->addr_len) &&
1811             nla_put(skb, IFLA_PERM_ADDRESS, dev->addr_len, dev->perm_addr))
1812                 goto nla_put_failure;
1813
1814         rcu_read_lock();
1815         if (rtnl_fill_link_af(skb, dev, ext_filter_mask))
1816                 goto nla_put_failure_rcu;
1817         rcu_read_unlock();
1818
1819         if (rtnl_fill_prop_list(skb, dev))
1820                 goto nla_put_failure;
1821
1822         nlmsg_end(skb, nlh);
1823         return 0;
1824
1825 nla_put_failure_rcu:
1826         rcu_read_unlock();
1827 nla_put_failure:
1828         nlmsg_cancel(skb, nlh);
1829         return -EMSGSIZE;
1830 }
1831
1832 static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
1833         [IFLA_IFNAME]           = { .type = NLA_STRING, .len = IFNAMSIZ-1 },
1834         [IFLA_ADDRESS]          = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1835         [IFLA_BROADCAST]        = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1836         [IFLA_MAP]              = { .len = sizeof(struct rtnl_link_ifmap) },
1837         [IFLA_MTU]              = { .type = NLA_U32 },
1838         [IFLA_LINK]             = { .type = NLA_U32 },
1839         [IFLA_MASTER]           = { .type = NLA_U32 },
1840         [IFLA_CARRIER]          = { .type = NLA_U8 },
1841         [IFLA_TXQLEN]           = { .type = NLA_U32 },
1842         [IFLA_WEIGHT]           = { .type = NLA_U32 },
1843         [IFLA_OPERSTATE]        = { .type = NLA_U8 },
1844         [IFLA_LINKMODE]         = { .type = NLA_U8 },
1845         [IFLA_LINKINFO]         = { .type = NLA_NESTED },
1846         [IFLA_NET_NS_PID]       = { .type = NLA_U32 },
1847         [IFLA_NET_NS_FD]        = { .type = NLA_U32 },
1848         /* IFLA_IFALIAS is a string, but policy is set to NLA_BINARY to
1849          * allow 0-length string (needed to remove an alias).
1850          */
1851         [IFLA_IFALIAS]          = { .type = NLA_BINARY, .len = IFALIASZ - 1 },
1852         [IFLA_VFINFO_LIST]      = {. type = NLA_NESTED },
1853         [IFLA_VF_PORTS]         = { .type = NLA_NESTED },
1854         [IFLA_PORT_SELF]        = { .type = NLA_NESTED },
1855         [IFLA_AF_SPEC]          = { .type = NLA_NESTED },
1856         [IFLA_EXT_MASK]         = { .type = NLA_U32 },
1857         [IFLA_PROMISCUITY]      = { .type = NLA_U32 },
1858         [IFLA_NUM_TX_QUEUES]    = { .type = NLA_U32 },
1859         [IFLA_NUM_RX_QUEUES]    = { .type = NLA_U32 },
1860         [IFLA_GSO_MAX_SEGS]     = { .type = NLA_U32 },
1861         [IFLA_GSO_MAX_SIZE]     = { .type = NLA_U32 },
1862         [IFLA_PHYS_PORT_ID]     = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1863         [IFLA_CARRIER_CHANGES]  = { .type = NLA_U32 },  /* ignored */
1864         [IFLA_PHYS_SWITCH_ID]   = { .type = NLA_BINARY, .len = MAX_PHYS_ITEM_ID_LEN },
1865         [IFLA_LINK_NETNSID]     = { .type = NLA_S32 },
1866         [IFLA_PROTO_DOWN]       = { .type = NLA_U8 },
1867         [IFLA_XDP]              = { .type = NLA_NESTED },
1868         [IFLA_EVENT]            = { .type = NLA_U32 },
1869         [IFLA_GROUP]            = { .type = NLA_U32 },
1870         [IFLA_TARGET_NETNSID]   = { .type = NLA_S32 },
1871         [IFLA_CARRIER_UP_COUNT] = { .type = NLA_U32 },
1872         [IFLA_CARRIER_DOWN_COUNT] = { .type = NLA_U32 },
1873         [IFLA_MIN_MTU]          = { .type = NLA_U32 },
1874         [IFLA_MAX_MTU]          = { .type = NLA_U32 },
1875         [IFLA_PROP_LIST]        = { .type = NLA_NESTED },
1876         [IFLA_ALT_IFNAME]       = { .type = NLA_STRING,
1877                                     .len = ALTIFNAMSIZ - 1 },
1878         [IFLA_PERM_ADDRESS]     = { .type = NLA_REJECT },
1879         [IFLA_PROTO_DOWN_REASON] = { .type = NLA_NESTED },
1880 };
1881
1882 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
1883         [IFLA_INFO_KIND]        = { .type = NLA_STRING },
1884         [IFLA_INFO_DATA]        = { .type = NLA_NESTED },
1885         [IFLA_INFO_SLAVE_KIND]  = { .type = NLA_STRING },
1886         [IFLA_INFO_SLAVE_DATA]  = { .type = NLA_NESTED },
1887 };
1888
1889 static const struct nla_policy ifla_vf_policy[IFLA_VF_MAX+1] = {
1890         [IFLA_VF_MAC]           = { .len = sizeof(struct ifla_vf_mac) },
1891         [IFLA_VF_BROADCAST]     = { .type = NLA_REJECT },
1892         [IFLA_VF_VLAN]          = { .len = sizeof(struct ifla_vf_vlan) },
1893         [IFLA_VF_VLAN_LIST]     = { .type = NLA_NESTED },
1894         [IFLA_VF_TX_RATE]       = { .len = sizeof(struct ifla_vf_tx_rate) },
1895         [IFLA_VF_SPOOFCHK]      = { .len = sizeof(struct ifla_vf_spoofchk) },
1896         [IFLA_VF_RATE]          = { .len = sizeof(struct ifla_vf_rate) },
1897         [IFLA_VF_LINK_STATE]    = { .len = sizeof(struct ifla_vf_link_state) },
1898         [IFLA_VF_RSS_QUERY_EN]  = { .len = sizeof(struct ifla_vf_rss_query_en) },
1899         [IFLA_VF_STATS]         = { .type = NLA_NESTED },
1900         [IFLA_VF_TRUST]         = { .len = sizeof(struct ifla_vf_trust) },
1901         [IFLA_VF_IB_NODE_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
1902         [IFLA_VF_IB_PORT_GUID]  = { .len = sizeof(struct ifla_vf_guid) },
1903 };
1904
1905 static const struct nla_policy ifla_port_policy[IFLA_PORT_MAX+1] = {
1906         [IFLA_PORT_VF]          = { .type = NLA_U32 },
1907         [IFLA_PORT_PROFILE]     = { .type = NLA_STRING,
1908                                     .len = PORT_PROFILE_MAX },
1909         [IFLA_PORT_INSTANCE_UUID] = { .type = NLA_BINARY,
1910                                       .len = PORT_UUID_MAX },
1911         [IFLA_PORT_HOST_UUID]   = { .type = NLA_STRING,
1912                                     .len = PORT_UUID_MAX },
1913         [IFLA_PORT_REQUEST]     = { .type = NLA_U8, },
1914         [IFLA_PORT_RESPONSE]    = { .type = NLA_U16, },
1915
1916         /* Unused, but we need to keep it here since user space could
1917          * fill it. It's also broken with regard to NLA_BINARY use in
1918          * combination with structs.
1919          */
1920         [IFLA_PORT_VSI_TYPE]    = { .type = NLA_BINARY,
1921                                     .len = sizeof(struct ifla_port_vsi) },
1922 };
1923
1924 static const struct nla_policy ifla_xdp_policy[IFLA_XDP_MAX + 1] = {
1925         [IFLA_XDP_UNSPEC]       = { .strict_start_type = IFLA_XDP_EXPECTED_FD },
1926         [IFLA_XDP_FD]           = { .type = NLA_S32 },
1927         [IFLA_XDP_EXPECTED_FD]  = { .type = NLA_S32 },
1928         [IFLA_XDP_ATTACHED]     = { .type = NLA_U8 },
1929         [IFLA_XDP_FLAGS]        = { .type = NLA_U32 },
1930         [IFLA_XDP_PROG_ID]      = { .type = NLA_U32 },
1931 };
1932
1933 static const struct rtnl_link_ops *linkinfo_to_kind_ops(const struct nlattr *nla)
1934 {
1935         const struct rtnl_link_ops *ops = NULL;
1936         struct nlattr *linfo[IFLA_INFO_MAX + 1];
1937
1938         if (nla_parse_nested_deprecated(linfo, IFLA_INFO_MAX, nla, ifla_info_policy, NULL) < 0)
1939                 return NULL;
1940
1941         if (linfo[IFLA_INFO_KIND]) {
1942                 char kind[MODULE_NAME_LEN];
1943
1944                 nla_strlcpy(kind, linfo[IFLA_INFO_KIND], sizeof(kind));
1945                 ops = rtnl_link_ops_get(kind);
1946         }
1947
1948         return ops;
1949 }
1950
1951 static bool link_master_filtered(struct net_device *dev, int master_idx)
1952 {
1953         struct net_device *master;
1954
1955         if (!master_idx)
1956                 return false;
1957
1958         master = netdev_master_upper_dev_get(dev);
1959         if (!master || master->ifindex != master_idx)
1960                 return true;
1961
1962         return false;
1963 }
1964
1965 static bool link_kind_filtered(const struct net_device *dev,
1966                                const struct rtnl_link_ops *kind_ops)
1967 {
1968         if (kind_ops && dev->rtnl_link_ops != kind_ops)
1969                 return true;
1970
1971         return false;
1972 }
1973
1974 static bool link_dump_filtered(struct net_device *dev,
1975                                int master_idx,
1976                                const struct rtnl_link_ops *kind_ops)
1977 {
1978         if (link_master_filtered(dev, master_idx) ||
1979             link_kind_filtered(dev, kind_ops))
1980                 return true;
1981
1982         return false;
1983 }
1984
1985 /**
1986  * rtnl_get_net_ns_capable - Get netns if sufficiently privileged.
1987  * @sk: netlink socket
1988  * @netnsid: network namespace identifier
1989  *
1990  * Returns the network namespace identified by netnsid on success or an error
1991  * pointer on failure.
1992  */
1993 struct net *rtnl_get_net_ns_capable(struct sock *sk, int netnsid)
1994 {
1995         struct net *net;
1996
1997         net = get_net_ns_by_id(sock_net(sk), netnsid);
1998         if (!net)
1999                 return ERR_PTR(-EINVAL);
2000
2001         /* For now, the caller is required to have CAP_NET_ADMIN in
2002          * the user namespace owning the target net ns.
2003          */
2004         if (!sk_ns_capable(sk, net->user_ns, CAP_NET_ADMIN)) {
2005                 put_net(net);
2006                 return ERR_PTR(-EACCES);
2007         }
2008         return net;
2009 }
2010 EXPORT_SYMBOL_GPL(rtnl_get_net_ns_capable);
2011
2012 static int rtnl_valid_dump_ifinfo_req(const struct nlmsghdr *nlh,
2013                                       bool strict_check, struct nlattr **tb,
2014                                       struct netlink_ext_ack *extack)
2015 {
2016         int hdrlen;
2017
2018         if (strict_check) {
2019                 struct ifinfomsg *ifm;
2020
2021                 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
2022                         NL_SET_ERR_MSG(extack, "Invalid header for link dump");
2023                         return -EINVAL;
2024                 }
2025
2026                 ifm = nlmsg_data(nlh);
2027                 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
2028                     ifm->ifi_change) {
2029                         NL_SET_ERR_MSG(extack, "Invalid values in header for link dump request");
2030                         return -EINVAL;
2031                 }
2032                 if (ifm->ifi_index) {
2033                         NL_SET_ERR_MSG(extack, "Filter by device index not supported for link dumps");
2034                         return -EINVAL;
2035                 }
2036
2037                 return nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb,
2038                                                      IFLA_MAX, ifla_policy,
2039                                                      extack);
2040         }
2041
2042         /* A hack to preserve kernel<->userspace interface.
2043          * The correct header is ifinfomsg. It is consistent with rtnl_getlink.
2044          * However, before Linux v3.9 the code here assumed rtgenmsg and that's
2045          * what iproute2 < v3.9.0 used.
2046          * We can detect the old iproute2. Even including the IFLA_EXT_MASK
2047          * attribute, its netlink message is shorter than struct ifinfomsg.
2048          */
2049         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
2050                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
2051
2052         return nlmsg_parse_deprecated(nlh, hdrlen, tb, IFLA_MAX, ifla_policy,
2053                                       extack);
2054 }
2055
2056 static int rtnl_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
2057 {
2058         struct netlink_ext_ack *extack = cb->extack;
2059         const struct nlmsghdr *nlh = cb->nlh;
2060         struct net *net = sock_net(skb->sk);
2061         struct net *tgt_net = net;
2062         int h, s_h;
2063         int idx = 0, s_idx;
2064         struct net_device *dev;
2065         struct hlist_head *head;
2066         struct nlattr *tb[IFLA_MAX+1];
2067         u32 ext_filter_mask = 0;
2068         const struct rtnl_link_ops *kind_ops = NULL;
2069         unsigned int flags = NLM_F_MULTI;
2070         int master_idx = 0;
2071         int netnsid = -1;
2072         int err, i;
2073
2074         s_h = cb->args[0];
2075         s_idx = cb->args[1];
2076
2077         err = rtnl_valid_dump_ifinfo_req(nlh, cb->strict_check, tb, extack);
2078         if (err < 0) {
2079                 if (cb->strict_check)
2080                         return err;
2081
2082                 goto walk_entries;
2083         }
2084
2085         for (i = 0; i <= IFLA_MAX; ++i) {
2086                 if (!tb[i])
2087                         continue;
2088
2089                 /* new attributes should only be added with strict checking */
2090                 switch (i) {
2091                 case IFLA_TARGET_NETNSID:
2092                         netnsid = nla_get_s32(tb[i]);
2093                         tgt_net = rtnl_get_net_ns_capable(skb->sk, netnsid);
2094                         if (IS_ERR(tgt_net)) {
2095                                 NL_SET_ERR_MSG(extack, "Invalid target network namespace id");
2096                                 return PTR_ERR(tgt_net);
2097                         }
2098                         break;
2099                 case IFLA_EXT_MASK:
2100                         ext_filter_mask = nla_get_u32(tb[i]);
2101                         break;
2102                 case IFLA_MASTER:
2103                         master_idx = nla_get_u32(tb[i]);
2104                         break;
2105                 case IFLA_LINKINFO:
2106                         kind_ops = linkinfo_to_kind_ops(tb[i]);
2107                         break;
2108                 default:
2109                         if (cb->strict_check) {
2110                                 NL_SET_ERR_MSG(extack, "Unsupported attribute in link dump request");
2111                                 return -EINVAL;
2112                         }
2113                 }
2114         }
2115
2116         if (master_idx || kind_ops)
2117                 flags |= NLM_F_DUMP_FILTERED;
2118
2119 walk_entries:
2120         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
2121                 idx = 0;
2122                 head = &tgt_net->dev_index_head[h];
2123                 hlist_for_each_entry(dev, head, index_hlist) {
2124                         if (link_dump_filtered(dev, master_idx, kind_ops))
2125                                 goto cont;
2126                         if (idx < s_idx)
2127                                 goto cont;
2128                         err = rtnl_fill_ifinfo(skb, dev, net,
2129                                                RTM_NEWLINK,
2130                                                NETLINK_CB(cb->skb).portid,
2131                                                nlh->nlmsg_seq, 0, flags,
2132                                                ext_filter_mask, 0, NULL, 0,
2133                                                netnsid, GFP_KERNEL);
2134
2135                         if (err < 0) {
2136                                 if (likely(skb->len))
2137                                         goto out;
2138
2139                                 goto out_err;
2140                         }
2141 cont:
2142                         idx++;
2143                 }
2144         }
2145 out:
2146         err = skb->len;
2147 out_err:
2148         cb->args[1] = idx;
2149         cb->args[0] = h;
2150         cb->seq = net->dev_base_seq;
2151         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
2152         if (netnsid >= 0)
2153                 put_net(tgt_net);
2154
2155         return err;
2156 }
2157
2158 int rtnl_nla_parse_ifla(struct nlattr **tb, const struct nlattr *head, int len,
2159                         struct netlink_ext_ack *exterr)
2160 {
2161         return nla_parse_deprecated(tb, IFLA_MAX, head, len, ifla_policy,
2162                                     exterr);
2163 }
2164 EXPORT_SYMBOL(rtnl_nla_parse_ifla);
2165
2166 struct net *rtnl_link_get_net(struct net *src_net, struct nlattr *tb[])
2167 {
2168         struct net *net;
2169         /* Examine the link attributes and figure out which
2170          * network namespace we are talking about.
2171          */
2172         if (tb[IFLA_NET_NS_PID])
2173                 net = get_net_ns_by_pid(nla_get_u32(tb[IFLA_NET_NS_PID]));
2174         else if (tb[IFLA_NET_NS_FD])
2175                 net = get_net_ns_by_fd(nla_get_u32(tb[IFLA_NET_NS_FD]));
2176         else
2177                 net = get_net(src_net);
2178         return net;
2179 }
2180 EXPORT_SYMBOL(rtnl_link_get_net);
2181
2182 /* Figure out which network namespace we are talking about by
2183  * examining the link attributes in the following order:
2184  *
2185  * 1. IFLA_NET_NS_PID
2186  * 2. IFLA_NET_NS_FD
2187  * 3. IFLA_TARGET_NETNSID
2188  */
2189 static struct net *rtnl_link_get_net_by_nlattr(struct net *src_net,
2190                                                struct nlattr *tb[])
2191 {
2192         struct net *net;
2193
2194         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD])
2195                 return rtnl_link_get_net(src_net, tb);
2196
2197         if (!tb[IFLA_TARGET_NETNSID])
2198                 return get_net(src_net);
2199
2200         net = get_net_ns_by_id(src_net, nla_get_u32(tb[IFLA_TARGET_NETNSID]));
2201         if (!net)
2202                 return ERR_PTR(-EINVAL);
2203
2204         return net;
2205 }
2206
2207 static struct net *rtnl_link_get_net_capable(const struct sk_buff *skb,
2208                                              struct net *src_net,
2209                                              struct nlattr *tb[], int cap)
2210 {
2211         struct net *net;
2212
2213         net = rtnl_link_get_net_by_nlattr(src_net, tb);
2214         if (IS_ERR(net))
2215                 return net;
2216
2217         if (!netlink_ns_capable(skb, net->user_ns, cap)) {
2218                 put_net(net);
2219                 return ERR_PTR(-EPERM);
2220         }
2221
2222         return net;
2223 }
2224
2225 /* Verify that rtnetlink requests do not pass additional properties
2226  * potentially referring to different network namespaces.
2227  */
2228 static int rtnl_ensure_unique_netns(struct nlattr *tb[],
2229                                     struct netlink_ext_ack *extack,
2230                                     bool netns_id_only)
2231 {
2232
2233         if (netns_id_only) {
2234                 if (!tb[IFLA_NET_NS_PID] && !tb[IFLA_NET_NS_FD])
2235                         return 0;
2236
2237                 NL_SET_ERR_MSG(extack, "specified netns attribute not supported");
2238                 return -EOPNOTSUPP;
2239         }
2240
2241         if (tb[IFLA_TARGET_NETNSID] && (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD]))
2242                 goto invalid_attr;
2243
2244         if (tb[IFLA_NET_NS_PID] && (tb[IFLA_TARGET_NETNSID] || tb[IFLA_NET_NS_FD]))
2245                 goto invalid_attr;
2246
2247         if (tb[IFLA_NET_NS_FD] && (tb[IFLA_TARGET_NETNSID] || tb[IFLA_NET_NS_PID]))
2248                 goto invalid_attr;
2249
2250         return 0;
2251
2252 invalid_attr:
2253         NL_SET_ERR_MSG(extack, "multiple netns identifying attributes specified");
2254         return -EINVAL;
2255 }
2256
2257 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
2258 {
2259         if (dev) {
2260                 if (tb[IFLA_ADDRESS] &&
2261                     nla_len(tb[IFLA_ADDRESS]) < dev->addr_len)
2262                         return -EINVAL;
2263
2264                 if (tb[IFLA_BROADCAST] &&
2265                     nla_len(tb[IFLA_BROADCAST]) < dev->addr_len)
2266                         return -EINVAL;
2267         }
2268
2269         if (tb[IFLA_AF_SPEC]) {
2270                 struct nlattr *af;
2271                 int rem, err;
2272
2273                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2274                         const struct rtnl_af_ops *af_ops;
2275
2276                         rcu_read_lock();
2277                         af_ops = rtnl_af_lookup(nla_type(af));
2278                         if (!af_ops) {
2279                                 rcu_read_unlock();
2280                                 return -EAFNOSUPPORT;
2281                         }
2282
2283                         if (!af_ops->set_link_af) {
2284                                 rcu_read_unlock();
2285                                 return -EOPNOTSUPP;
2286                         }
2287
2288                         if (af_ops->validate_link_af) {
2289                                 err = af_ops->validate_link_af(dev, af);
2290                                 if (err < 0) {
2291                                         rcu_read_unlock();
2292                                         return err;
2293                                 }
2294                         }
2295
2296                         rcu_read_unlock();
2297                 }
2298         }
2299
2300         return 0;
2301 }
2302
2303 static int handle_infiniband_guid(struct net_device *dev, struct ifla_vf_guid *ivt,
2304                                   int guid_type)
2305 {
2306         const struct net_device_ops *ops = dev->netdev_ops;
2307
2308         return ops->ndo_set_vf_guid(dev, ivt->vf, ivt->guid, guid_type);
2309 }
2310
2311 static int handle_vf_guid(struct net_device *dev, struct ifla_vf_guid *ivt, int guid_type)
2312 {
2313         if (dev->type != ARPHRD_INFINIBAND)
2314                 return -EOPNOTSUPP;
2315
2316         return handle_infiniband_guid(dev, ivt, guid_type);
2317 }
2318
2319 static int do_setvfinfo(struct net_device *dev, struct nlattr **tb)
2320 {
2321         const struct net_device_ops *ops = dev->netdev_ops;
2322         int err = -EINVAL;
2323
2324         if (tb[IFLA_VF_MAC]) {
2325                 struct ifla_vf_mac *ivm = nla_data(tb[IFLA_VF_MAC]);
2326
2327                 if (ivm->vf >= INT_MAX)
2328                         return -EINVAL;
2329                 err = -EOPNOTSUPP;
2330                 if (ops->ndo_set_vf_mac)
2331                         err = ops->ndo_set_vf_mac(dev, ivm->vf,
2332                                                   ivm->mac);
2333                 if (err < 0)
2334                         return err;
2335         }
2336
2337         if (tb[IFLA_VF_VLAN]) {
2338                 struct ifla_vf_vlan *ivv = nla_data(tb[IFLA_VF_VLAN]);
2339
2340                 if (ivv->vf >= INT_MAX)
2341                         return -EINVAL;
2342                 err = -EOPNOTSUPP;
2343                 if (ops->ndo_set_vf_vlan)
2344                         err = ops->ndo_set_vf_vlan(dev, ivv->vf, ivv->vlan,
2345                                                    ivv->qos,
2346                                                    htons(ETH_P_8021Q));
2347                 if (err < 0)
2348                         return err;
2349         }
2350
2351         if (tb[IFLA_VF_VLAN_LIST]) {
2352                 struct ifla_vf_vlan_info *ivvl[MAX_VLAN_LIST_LEN];
2353                 struct nlattr *attr;
2354                 int rem, len = 0;
2355
2356                 err = -EOPNOTSUPP;
2357                 if (!ops->ndo_set_vf_vlan)
2358                         return err;
2359
2360                 nla_for_each_nested(attr, tb[IFLA_VF_VLAN_LIST], rem) {
2361                         if (nla_type(attr) != IFLA_VF_VLAN_INFO ||
2362                             nla_len(attr) < NLA_HDRLEN) {
2363                                 return -EINVAL;
2364                         }
2365                         if (len >= MAX_VLAN_LIST_LEN)
2366                                 return -EOPNOTSUPP;
2367                         ivvl[len] = nla_data(attr);
2368
2369                         len++;
2370                 }
2371                 if (len == 0)
2372                         return -EINVAL;
2373
2374                 if (ivvl[0]->vf >= INT_MAX)
2375                         return -EINVAL;
2376                 err = ops->ndo_set_vf_vlan(dev, ivvl[0]->vf, ivvl[0]->vlan,
2377                                            ivvl[0]->qos, ivvl[0]->vlan_proto);
2378                 if (err < 0)
2379                         return err;
2380         }
2381
2382         if (tb[IFLA_VF_TX_RATE]) {
2383                 struct ifla_vf_tx_rate *ivt = nla_data(tb[IFLA_VF_TX_RATE]);
2384                 struct ifla_vf_info ivf;
2385
2386                 if (ivt->vf >= INT_MAX)
2387                         return -EINVAL;
2388                 err = -EOPNOTSUPP;
2389                 if (ops->ndo_get_vf_config)
2390                         err = ops->ndo_get_vf_config(dev, ivt->vf, &ivf);
2391                 if (err < 0)
2392                         return err;
2393
2394                 err = -EOPNOTSUPP;
2395                 if (ops->ndo_set_vf_rate)
2396                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
2397                                                    ivf.min_tx_rate,
2398                                                    ivt->rate);
2399                 if (err < 0)
2400                         return err;
2401         }
2402
2403         if (tb[IFLA_VF_RATE]) {
2404                 struct ifla_vf_rate *ivt = nla_data(tb[IFLA_VF_RATE]);
2405
2406                 if (ivt->vf >= INT_MAX)
2407                         return -EINVAL;
2408                 err = -EOPNOTSUPP;
2409                 if (ops->ndo_set_vf_rate)
2410                         err = ops->ndo_set_vf_rate(dev, ivt->vf,
2411                                                    ivt->min_tx_rate,
2412                                                    ivt->max_tx_rate);
2413                 if (err < 0)
2414                         return err;
2415         }
2416
2417         if (tb[IFLA_VF_SPOOFCHK]) {
2418                 struct ifla_vf_spoofchk *ivs = nla_data(tb[IFLA_VF_SPOOFCHK]);
2419
2420                 if (ivs->vf >= INT_MAX)
2421                         return -EINVAL;
2422                 err = -EOPNOTSUPP;
2423                 if (ops->ndo_set_vf_spoofchk)
2424                         err = ops->ndo_set_vf_spoofchk(dev, ivs->vf,
2425                                                        ivs->setting);
2426                 if (err < 0)
2427                         return err;
2428         }
2429
2430         if (tb[IFLA_VF_LINK_STATE]) {
2431                 struct ifla_vf_link_state *ivl = nla_data(tb[IFLA_VF_LINK_STATE]);
2432
2433                 if (ivl->vf >= INT_MAX)
2434                         return -EINVAL;
2435                 err = -EOPNOTSUPP;
2436                 if (ops->ndo_set_vf_link_state)
2437                         err = ops->ndo_set_vf_link_state(dev, ivl->vf,
2438                                                          ivl->link_state);
2439                 if (err < 0)
2440                         return err;
2441         }
2442
2443         if (tb[IFLA_VF_RSS_QUERY_EN]) {
2444                 struct ifla_vf_rss_query_en *ivrssq_en;
2445
2446                 err = -EOPNOTSUPP;
2447                 ivrssq_en = nla_data(tb[IFLA_VF_RSS_QUERY_EN]);
2448                 if (ivrssq_en->vf >= INT_MAX)
2449                         return -EINVAL;
2450                 if (ops->ndo_set_vf_rss_query_en)
2451                         err = ops->ndo_set_vf_rss_query_en(dev, ivrssq_en->vf,
2452                                                            ivrssq_en->setting);
2453                 if (err < 0)
2454                         return err;
2455         }
2456
2457         if (tb[IFLA_VF_TRUST]) {
2458                 struct ifla_vf_trust *ivt = nla_data(tb[IFLA_VF_TRUST]);
2459
2460                 if (ivt->vf >= INT_MAX)
2461                         return -EINVAL;
2462                 err = -EOPNOTSUPP;
2463                 if (ops->ndo_set_vf_trust)
2464                         err = ops->ndo_set_vf_trust(dev, ivt->vf, ivt->setting);
2465                 if (err < 0)
2466                         return err;
2467         }
2468
2469         if (tb[IFLA_VF_IB_NODE_GUID]) {
2470                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_NODE_GUID]);
2471
2472                 if (ivt->vf >= INT_MAX)
2473                         return -EINVAL;
2474                 if (!ops->ndo_set_vf_guid)
2475                         return -EOPNOTSUPP;
2476                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_NODE_GUID);
2477         }
2478
2479         if (tb[IFLA_VF_IB_PORT_GUID]) {
2480                 struct ifla_vf_guid *ivt = nla_data(tb[IFLA_VF_IB_PORT_GUID]);
2481
2482                 if (ivt->vf >= INT_MAX)
2483                         return -EINVAL;
2484                 if (!ops->ndo_set_vf_guid)
2485                         return -EOPNOTSUPP;
2486
2487                 return handle_vf_guid(dev, ivt, IFLA_VF_IB_PORT_GUID);
2488         }
2489
2490         return err;
2491 }
2492
2493 static int do_set_master(struct net_device *dev, int ifindex,
2494                          struct netlink_ext_ack *extack)
2495 {
2496         struct net_device *upper_dev = netdev_master_upper_dev_get(dev);
2497         const struct net_device_ops *ops;
2498         int err;
2499
2500         if (upper_dev) {
2501                 if (upper_dev->ifindex == ifindex)
2502                         return 0;
2503                 ops = upper_dev->netdev_ops;
2504                 if (ops->ndo_del_slave) {
2505                         err = ops->ndo_del_slave(upper_dev, dev);
2506                         if (err)
2507                                 return err;
2508                 } else {
2509                         return -EOPNOTSUPP;
2510                 }
2511         }
2512
2513         if (ifindex) {
2514                 upper_dev = __dev_get_by_index(dev_net(dev), ifindex);
2515                 if (!upper_dev)
2516                         return -EINVAL;
2517                 ops = upper_dev->netdev_ops;
2518                 if (ops->ndo_add_slave) {
2519                         err = ops->ndo_add_slave(upper_dev, dev, extack);
2520                         if (err)
2521                                 return err;
2522                 } else {
2523                         return -EOPNOTSUPP;
2524                 }
2525         }
2526         return 0;
2527 }
2528
2529 static const struct nla_policy ifla_proto_down_reason_policy[IFLA_PROTO_DOWN_REASON_VALUE + 1] = {
2530         [IFLA_PROTO_DOWN_REASON_MASK]   = { .type = NLA_U32 },
2531         [IFLA_PROTO_DOWN_REASON_VALUE]  = { .type = NLA_U32 },
2532 };
2533
2534 static int do_set_proto_down(struct net_device *dev,
2535                              struct nlattr *nl_proto_down,
2536                              struct nlattr *nl_proto_down_reason,
2537                              struct netlink_ext_ack *extack)
2538 {
2539         struct nlattr *pdreason[IFLA_PROTO_DOWN_REASON_MAX + 1];
2540         const struct net_device_ops *ops = dev->netdev_ops;
2541         unsigned long mask = 0;
2542         u32 value;
2543         bool proto_down;
2544         int err;
2545
2546         if (!ops->ndo_change_proto_down) {
2547                 NL_SET_ERR_MSG(extack,  "Protodown not supported by device");
2548                 return -EOPNOTSUPP;
2549         }
2550
2551         if (nl_proto_down_reason) {
2552                 err = nla_parse_nested_deprecated(pdreason,
2553                                                   IFLA_PROTO_DOWN_REASON_MAX,
2554                                                   nl_proto_down_reason,
2555                                                   ifla_proto_down_reason_policy,
2556                                                   NULL);
2557                 if (err < 0)
2558                         return err;
2559
2560                 if (!pdreason[IFLA_PROTO_DOWN_REASON_VALUE]) {
2561                         NL_SET_ERR_MSG(extack, "Invalid protodown reason value");
2562                         return -EINVAL;
2563                 }
2564
2565                 value = nla_get_u32(pdreason[IFLA_PROTO_DOWN_REASON_VALUE]);
2566
2567                 if (pdreason[IFLA_PROTO_DOWN_REASON_MASK])
2568                         mask = nla_get_u32(pdreason[IFLA_PROTO_DOWN_REASON_MASK]);
2569
2570                 dev_change_proto_down_reason(dev, mask, value);
2571         }
2572
2573         if (nl_proto_down) {
2574                 proto_down = nla_get_u8(nl_proto_down);
2575
2576                 /* Dont turn off protodown if there are active reasons */
2577                 if (!proto_down && dev->proto_down_reason) {
2578                         NL_SET_ERR_MSG(extack, "Cannot clear protodown, active reasons");
2579                         return -EBUSY;
2580                 }
2581                 err = dev_change_proto_down(dev,
2582                                             proto_down);
2583                 if (err)
2584                         return err;
2585         }
2586
2587         return 0;
2588 }
2589
2590 #define DO_SETLINK_MODIFIED     0x01
2591 /* notify flag means notify + modified. */
2592 #define DO_SETLINK_NOTIFY       0x03
2593 static int do_setlink(const struct sk_buff *skb,
2594                       struct net_device *dev, struct ifinfomsg *ifm,
2595                       struct netlink_ext_ack *extack,
2596                       struct nlattr **tb, char *ifname, int status)
2597 {
2598         const struct net_device_ops *ops = dev->netdev_ops;
2599         int err;
2600
2601         err = validate_linkmsg(dev, tb);
2602         if (err < 0)
2603                 return err;
2604
2605         if (tb[IFLA_NET_NS_PID] || tb[IFLA_NET_NS_FD] || tb[IFLA_TARGET_NETNSID]) {
2606                 const char *pat = ifname && ifname[0] ? ifname : NULL;
2607                 struct net *net = rtnl_link_get_net_capable(skb, dev_net(dev),
2608                                                             tb, CAP_NET_ADMIN);
2609                 if (IS_ERR(net)) {
2610                         err = PTR_ERR(net);
2611                         goto errout;
2612                 }
2613
2614                 err = dev_change_net_namespace(dev, net, pat);
2615                 put_net(net);
2616                 if (err)
2617                         goto errout;
2618                 status |= DO_SETLINK_MODIFIED;
2619         }
2620
2621         if (tb[IFLA_MAP]) {
2622                 struct rtnl_link_ifmap *u_map;
2623                 struct ifmap k_map;
2624
2625                 if (!ops->ndo_set_config) {
2626                         err = -EOPNOTSUPP;
2627                         goto errout;
2628                 }
2629
2630                 if (!netif_device_present(dev)) {
2631                         err = -ENODEV;
2632                         goto errout;
2633                 }
2634
2635                 u_map = nla_data(tb[IFLA_MAP]);
2636                 k_map.mem_start = (unsigned long) u_map->mem_start;
2637                 k_map.mem_end = (unsigned long) u_map->mem_end;
2638                 k_map.base_addr = (unsigned short) u_map->base_addr;
2639                 k_map.irq = (unsigned char) u_map->irq;
2640                 k_map.dma = (unsigned char) u_map->dma;
2641                 k_map.port = (unsigned char) u_map->port;
2642
2643                 err = ops->ndo_set_config(dev, &k_map);
2644                 if (err < 0)
2645                         goto errout;
2646
2647                 status |= DO_SETLINK_NOTIFY;
2648         }
2649
2650         if (tb[IFLA_ADDRESS]) {
2651                 struct sockaddr *sa;
2652                 int len;
2653
2654                 len = sizeof(sa_family_t) + max_t(size_t, dev->addr_len,
2655                                                   sizeof(*sa));
2656                 sa = kmalloc(len, GFP_KERNEL);
2657                 if (!sa) {
2658                         err = -ENOMEM;
2659                         goto errout;
2660                 }
2661                 sa->sa_family = dev->type;
2662                 memcpy(sa->sa_data, nla_data(tb[IFLA_ADDRESS]),
2663                        dev->addr_len);
2664                 err = dev_set_mac_address_user(dev, sa, extack);
2665                 kfree(sa);
2666                 if (err)
2667                         goto errout;
2668                 status |= DO_SETLINK_MODIFIED;
2669         }
2670
2671         if (tb[IFLA_MTU]) {
2672                 err = dev_set_mtu_ext(dev, nla_get_u32(tb[IFLA_MTU]), extack);
2673                 if (err < 0)
2674                         goto errout;
2675                 status |= DO_SETLINK_MODIFIED;
2676         }
2677
2678         if (tb[IFLA_GROUP]) {
2679                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
2680                 status |= DO_SETLINK_NOTIFY;
2681         }
2682
2683         /*
2684          * Interface selected by interface index but interface
2685          * name provided implies that a name change has been
2686          * requested.
2687          */
2688         if (ifm->ifi_index > 0 && ifname[0]) {
2689                 err = dev_change_name(dev, ifname);
2690                 if (err < 0)
2691                         goto errout;
2692                 status |= DO_SETLINK_MODIFIED;
2693         }
2694
2695         if (tb[IFLA_IFALIAS]) {
2696                 err = dev_set_alias(dev, nla_data(tb[IFLA_IFALIAS]),
2697                                     nla_len(tb[IFLA_IFALIAS]));
2698                 if (err < 0)
2699                         goto errout;
2700                 status |= DO_SETLINK_NOTIFY;
2701         }
2702
2703         if (tb[IFLA_BROADCAST]) {
2704                 nla_memcpy(dev->broadcast, tb[IFLA_BROADCAST], dev->addr_len);
2705                 call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
2706         }
2707
2708         if (ifm->ifi_flags || ifm->ifi_change) {
2709                 err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
2710                                        extack);
2711                 if (err < 0)
2712                         goto errout;
2713         }
2714
2715         if (tb[IFLA_MASTER]) {
2716                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
2717                 if (err)
2718                         goto errout;
2719                 status |= DO_SETLINK_MODIFIED;
2720         }
2721
2722         if (tb[IFLA_CARRIER]) {
2723                 err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
2724                 if (err)
2725                         goto errout;
2726                 status |= DO_SETLINK_MODIFIED;
2727         }
2728
2729         if (tb[IFLA_TXQLEN]) {
2730                 unsigned int value = nla_get_u32(tb[IFLA_TXQLEN]);
2731
2732                 err = dev_change_tx_queue_len(dev, value);
2733                 if (err)
2734                         goto errout;
2735                 status |= DO_SETLINK_MODIFIED;
2736         }
2737
2738         if (tb[IFLA_GSO_MAX_SIZE]) {
2739                 u32 max_size = nla_get_u32(tb[IFLA_GSO_MAX_SIZE]);
2740
2741                 if (max_size > GSO_MAX_SIZE) {
2742                         err = -EINVAL;
2743                         goto errout;
2744                 }
2745
2746                 if (dev->gso_max_size ^ max_size) {
2747                         netif_set_gso_max_size(dev, max_size);
2748                         status |= DO_SETLINK_MODIFIED;
2749                 }
2750         }
2751
2752         if (tb[IFLA_GSO_MAX_SEGS]) {
2753                 u32 max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]);
2754
2755                 if (max_segs > GSO_MAX_SEGS) {
2756                         err = -EINVAL;
2757                         goto errout;
2758                 }
2759
2760                 if (dev->gso_max_segs ^ max_segs) {
2761                         dev->gso_max_segs = max_segs;
2762                         status |= DO_SETLINK_MODIFIED;
2763                 }
2764         }
2765
2766         if (tb[IFLA_OPERSTATE])
2767                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
2768
2769         if (tb[IFLA_LINKMODE]) {
2770                 unsigned char value = nla_get_u8(tb[IFLA_LINKMODE]);
2771
2772                 write_lock_bh(&dev_base_lock);
2773                 if (dev->link_mode ^ value)
2774                         status |= DO_SETLINK_NOTIFY;
2775                 dev->link_mode = value;
2776                 write_unlock_bh(&dev_base_lock);
2777         }
2778
2779         if (tb[IFLA_VFINFO_LIST]) {
2780                 struct nlattr *vfinfo[IFLA_VF_MAX + 1];
2781                 struct nlattr *attr;
2782                 int rem;
2783
2784                 nla_for_each_nested(attr, tb[IFLA_VFINFO_LIST], rem) {
2785                         if (nla_type(attr) != IFLA_VF_INFO ||
2786                             nla_len(attr) < NLA_HDRLEN) {
2787                                 err = -EINVAL;
2788                                 goto errout;
2789                         }
2790                         err = nla_parse_nested_deprecated(vfinfo, IFLA_VF_MAX,
2791                                                           attr,
2792                                                           ifla_vf_policy,
2793                                                           NULL);
2794                         if (err < 0)
2795                                 goto errout;
2796                         err = do_setvfinfo(dev, vfinfo);
2797                         if (err < 0)
2798                                 goto errout;
2799                         status |= DO_SETLINK_NOTIFY;
2800                 }
2801         }
2802         err = 0;
2803
2804         if (tb[IFLA_VF_PORTS]) {
2805                 struct nlattr *port[IFLA_PORT_MAX+1];
2806                 struct nlattr *attr;
2807                 int vf;
2808                 int rem;
2809
2810                 err = -EOPNOTSUPP;
2811                 if (!ops->ndo_set_vf_port)
2812                         goto errout;
2813
2814                 nla_for_each_nested(attr, tb[IFLA_VF_PORTS], rem) {
2815                         if (nla_type(attr) != IFLA_VF_PORT ||
2816                             nla_len(attr) < NLA_HDRLEN) {
2817                                 err = -EINVAL;
2818                                 goto errout;
2819                         }
2820                         err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX,
2821                                                           attr,
2822                                                           ifla_port_policy,
2823                                                           NULL);
2824                         if (err < 0)
2825                                 goto errout;
2826                         if (!port[IFLA_PORT_VF]) {
2827                                 err = -EOPNOTSUPP;
2828                                 goto errout;
2829                         }
2830                         vf = nla_get_u32(port[IFLA_PORT_VF]);
2831                         err = ops->ndo_set_vf_port(dev, vf, port);
2832                         if (err < 0)
2833                                 goto errout;
2834                         status |= DO_SETLINK_NOTIFY;
2835                 }
2836         }
2837         err = 0;
2838
2839         if (tb[IFLA_PORT_SELF]) {
2840                 struct nlattr *port[IFLA_PORT_MAX+1];
2841
2842                 err = nla_parse_nested_deprecated(port, IFLA_PORT_MAX,
2843                                                   tb[IFLA_PORT_SELF],
2844                                                   ifla_port_policy, NULL);
2845                 if (err < 0)
2846                         goto errout;
2847
2848                 err = -EOPNOTSUPP;
2849                 if (ops->ndo_set_vf_port)
2850                         err = ops->ndo_set_vf_port(dev, PORT_SELF_VF, port);
2851                 if (err < 0)
2852                         goto errout;
2853                 status |= DO_SETLINK_NOTIFY;
2854         }
2855
2856         if (tb[IFLA_AF_SPEC]) {
2857                 struct nlattr *af;
2858                 int rem;
2859
2860                 nla_for_each_nested(af, tb[IFLA_AF_SPEC], rem) {
2861                         const struct rtnl_af_ops *af_ops;
2862
2863                         rcu_read_lock();
2864
2865                         BUG_ON(!(af_ops = rtnl_af_lookup(nla_type(af))));
2866
2867                         err = af_ops->set_link_af(dev, af);
2868                         if (err < 0) {
2869                                 rcu_read_unlock();
2870                                 goto errout;
2871                         }
2872
2873                         rcu_read_unlock();
2874                         status |= DO_SETLINK_NOTIFY;
2875                 }
2876         }
2877         err = 0;
2878
2879         if (tb[IFLA_PROTO_DOWN] || tb[IFLA_PROTO_DOWN_REASON]) {
2880                 err = do_set_proto_down(dev, tb[IFLA_PROTO_DOWN],
2881                                         tb[IFLA_PROTO_DOWN_REASON], extack);
2882                 if (err)
2883                         goto errout;
2884                 status |= DO_SETLINK_NOTIFY;
2885         }
2886
2887         if (tb[IFLA_XDP]) {
2888                 struct nlattr *xdp[IFLA_XDP_MAX + 1];
2889                 u32 xdp_flags = 0;
2890
2891                 err = nla_parse_nested_deprecated(xdp, IFLA_XDP_MAX,
2892                                                   tb[IFLA_XDP],
2893                                                   ifla_xdp_policy, NULL);
2894                 if (err < 0)
2895                         goto errout;
2896
2897                 if (xdp[IFLA_XDP_ATTACHED] || xdp[IFLA_XDP_PROG_ID]) {
2898                         err = -EINVAL;
2899                         goto errout;
2900                 }
2901
2902                 if (xdp[IFLA_XDP_FLAGS]) {
2903                         xdp_flags = nla_get_u32(xdp[IFLA_XDP_FLAGS]);
2904                         if (xdp_flags & ~XDP_FLAGS_MASK) {
2905                                 err = -EINVAL;
2906                                 goto errout;
2907                         }
2908                         if (hweight32(xdp_flags & XDP_FLAGS_MODES) > 1) {
2909                                 err = -EINVAL;
2910                                 goto errout;
2911                         }
2912                 }
2913
2914                 if (xdp[IFLA_XDP_FD]) {
2915                         int expected_fd = -1;
2916
2917                         if (xdp_flags & XDP_FLAGS_REPLACE) {
2918                                 if (!xdp[IFLA_XDP_EXPECTED_FD]) {
2919                                         err = -EINVAL;
2920                                         goto errout;
2921                                 }
2922                                 expected_fd =
2923                                         nla_get_s32(xdp[IFLA_XDP_EXPECTED_FD]);
2924                         }
2925
2926                         err = dev_change_xdp_fd(dev, extack,
2927                                                 nla_get_s32(xdp[IFLA_XDP_FD]),
2928                                                 expected_fd,
2929                                                 xdp_flags);
2930                         if (err)
2931                                 goto errout;
2932                         status |= DO_SETLINK_NOTIFY;
2933                 }
2934         }
2935
2936 errout:
2937         if (status & DO_SETLINK_MODIFIED) {
2938                 if ((status & DO_SETLINK_NOTIFY) == DO_SETLINK_NOTIFY)
2939                         netdev_state_change(dev);
2940
2941                 if (err < 0)
2942                         net_warn_ratelimited("A link change request failed with some changes committed already. Interface %s may have been left with an inconsistent configuration, please check.\n",
2943                                              dev->name);
2944         }
2945
2946         return err;
2947 }
2948
2949 static struct net_device *rtnl_dev_get(struct net *net,
2950                                        struct nlattr *ifname_attr,
2951                                        struct nlattr *altifname_attr,
2952                                        char *ifname)
2953 {
2954         char buffer[ALTIFNAMSIZ];
2955
2956         if (!ifname) {
2957                 ifname = buffer;
2958                 if (ifname_attr)
2959                         nla_strlcpy(ifname, ifname_attr, IFNAMSIZ);
2960                 else if (altifname_attr)
2961                         nla_strlcpy(ifname, altifname_attr, ALTIFNAMSIZ);
2962                 else
2963                         return NULL;
2964         }
2965
2966         return __dev_get_by_name(net, ifname);
2967 }
2968
2969 static int rtnl_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
2970                         struct netlink_ext_ack *extack)
2971 {
2972         struct net *net = sock_net(skb->sk);
2973         struct ifinfomsg *ifm;
2974         struct net_device *dev;
2975         int err;
2976         struct nlattr *tb[IFLA_MAX+1];
2977         char ifname[IFNAMSIZ];
2978
2979         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
2980                                      ifla_policy, extack);
2981         if (err < 0)
2982                 goto errout;
2983
2984         err = rtnl_ensure_unique_netns(tb, extack, false);
2985         if (err < 0)
2986                 goto errout;
2987
2988         if (tb[IFLA_IFNAME])
2989                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
2990         else
2991                 ifname[0] = '\0';
2992
2993         err = -EINVAL;
2994         ifm = nlmsg_data(nlh);
2995         if (ifm->ifi_index > 0)
2996                 dev = __dev_get_by_index(net, ifm->ifi_index);
2997         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
2998                 dev = rtnl_dev_get(net, NULL, tb[IFLA_ALT_IFNAME], ifname);
2999         else
3000                 goto errout;
3001
3002         if (dev == NULL) {
3003                 err = -ENODEV;
3004                 goto errout;
3005         }
3006
3007         err = do_setlink(skb, dev, ifm, extack, tb, ifname, 0);
3008 errout:
3009         return err;
3010 }
3011
3012 static int rtnl_group_dellink(const struct net *net, int group)
3013 {
3014         struct net_device *dev, *aux;
3015         LIST_HEAD(list_kill);
3016         bool found = false;
3017
3018         if (!group)
3019                 return -EPERM;
3020
3021         for_each_netdev(net, dev) {
3022                 if (dev->group == group) {
3023                         const struct rtnl_link_ops *ops;
3024
3025                         found = true;
3026                         ops = dev->rtnl_link_ops;
3027                         if (!ops || !ops->dellink)
3028                                 return -EOPNOTSUPP;
3029                 }
3030         }
3031
3032         if (!found)
3033                 return -ENODEV;
3034
3035         for_each_netdev_safe(net, dev, aux) {
3036                 if (dev->group == group) {
3037                         const struct rtnl_link_ops *ops;
3038
3039                         ops = dev->rtnl_link_ops;
3040                         ops->dellink(dev, &list_kill);
3041                 }
3042         }
3043         unregister_netdevice_many(&list_kill);
3044
3045         return 0;
3046 }
3047
3048 int rtnl_delete_link(struct net_device *dev)
3049 {
3050         const struct rtnl_link_ops *ops;
3051         LIST_HEAD(list_kill);
3052
3053         ops = dev->rtnl_link_ops;
3054         if (!ops || !ops->dellink)
3055                 return -EOPNOTSUPP;
3056
3057         ops->dellink(dev, &list_kill);
3058         unregister_netdevice_many(&list_kill);
3059
3060         return 0;
3061 }
3062 EXPORT_SYMBOL_GPL(rtnl_delete_link);
3063
3064 static int rtnl_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
3065                         struct netlink_ext_ack *extack)
3066 {
3067         struct net *net = sock_net(skb->sk);
3068         struct net *tgt_net = net;
3069         struct net_device *dev = NULL;
3070         struct ifinfomsg *ifm;
3071         struct nlattr *tb[IFLA_MAX+1];
3072         int err;
3073         int netnsid = -1;
3074
3075         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3076                                      ifla_policy, extack);
3077         if (err < 0)
3078                 return err;
3079
3080         err = rtnl_ensure_unique_netns(tb, extack, true);
3081         if (err < 0)
3082                 return err;
3083
3084         if (tb[IFLA_TARGET_NETNSID]) {
3085                 netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
3086                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
3087                 if (IS_ERR(tgt_net))
3088                         return PTR_ERR(tgt_net);
3089         }
3090
3091         err = -EINVAL;
3092         ifm = nlmsg_data(nlh);
3093         if (ifm->ifi_index > 0)
3094                 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
3095         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3096                 dev = rtnl_dev_get(net, tb[IFLA_IFNAME],
3097                                    tb[IFLA_ALT_IFNAME], NULL);
3098         else if (tb[IFLA_GROUP])
3099                 err = rtnl_group_dellink(tgt_net, nla_get_u32(tb[IFLA_GROUP]));
3100         else
3101                 goto out;
3102
3103         if (!dev) {
3104                 if (tb[IFLA_IFNAME] || ifm->ifi_index > 0)
3105                         err = -ENODEV;
3106
3107                 goto out;
3108         }
3109
3110         err = rtnl_delete_link(dev);
3111
3112 out:
3113         if (netnsid >= 0)
3114                 put_net(tgt_net);
3115
3116         return err;
3117 }
3118
3119 int rtnl_configure_link(struct net_device *dev, const struct ifinfomsg *ifm)
3120 {
3121         unsigned int old_flags;
3122         int err;
3123
3124         old_flags = dev->flags;
3125         if (ifm && (ifm->ifi_flags || ifm->ifi_change)) {
3126                 err = __dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
3127                                          NULL);
3128                 if (err < 0)
3129                         return err;
3130         }
3131
3132         if (dev->rtnl_link_state == RTNL_LINK_INITIALIZED) {
3133                 __dev_notify_flags(dev, old_flags, (old_flags ^ dev->flags));
3134         } else {
3135                 dev->rtnl_link_state = RTNL_LINK_INITIALIZED;
3136                 __dev_notify_flags(dev, old_flags, ~0U);
3137         }
3138         return 0;
3139 }
3140 EXPORT_SYMBOL(rtnl_configure_link);
3141
3142 struct net_device *rtnl_create_link(struct net *net, const char *ifname,
3143                                     unsigned char name_assign_type,
3144                                     const struct rtnl_link_ops *ops,
3145                                     struct nlattr *tb[],
3146                                     struct netlink_ext_ack *extack)
3147 {
3148         struct net_device *dev;
3149         unsigned int num_tx_queues = 1;
3150         unsigned int num_rx_queues = 1;
3151
3152         if (tb[IFLA_NUM_TX_QUEUES])
3153                 num_tx_queues = nla_get_u32(tb[IFLA_NUM_TX_QUEUES]);
3154         else if (ops->get_num_tx_queues)
3155                 num_tx_queues = ops->get_num_tx_queues();
3156
3157         if (tb[IFLA_NUM_RX_QUEUES])
3158                 num_rx_queues = nla_get_u32(tb[IFLA_NUM_RX_QUEUES]);
3159         else if (ops->get_num_rx_queues)
3160                 num_rx_queues = ops->get_num_rx_queues();
3161
3162         if (num_tx_queues < 1 || num_tx_queues > 4096) {
3163                 NL_SET_ERR_MSG(extack, "Invalid number of transmit queues");
3164                 return ERR_PTR(-EINVAL);
3165         }
3166
3167         if (num_rx_queues < 1 || num_rx_queues > 4096) {
3168                 NL_SET_ERR_MSG(extack, "Invalid number of receive queues");
3169                 return ERR_PTR(-EINVAL);
3170         }
3171
3172         dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
3173                                ops->setup, num_tx_queues, num_rx_queues);
3174         if (!dev)
3175                 return ERR_PTR(-ENOMEM);
3176
3177         dev_net_set(dev, net);
3178         dev->rtnl_link_ops = ops;
3179         dev->rtnl_link_state = RTNL_LINK_INITIALIZING;
3180
3181         if (tb[IFLA_MTU]) {
3182                 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
3183                 int err;
3184
3185                 err = dev_validate_mtu(dev, mtu, extack);
3186                 if (err) {
3187                         free_netdev(dev);
3188                         return ERR_PTR(err);
3189                 }
3190                 dev->mtu = mtu;
3191         }
3192         if (tb[IFLA_ADDRESS]) {
3193                 memcpy(dev->dev_addr, nla_data(tb[IFLA_ADDRESS]),
3194                                 nla_len(tb[IFLA_ADDRESS]));
3195                 dev->addr_assign_type = NET_ADDR_SET;
3196         }
3197         if (tb[IFLA_BROADCAST])
3198                 memcpy(dev->broadcast, nla_data(tb[IFLA_BROADCAST]),
3199                                 nla_len(tb[IFLA_BROADCAST]));
3200         if (tb[IFLA_TXQLEN])
3201                 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
3202         if (tb[IFLA_OPERSTATE])
3203                 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
3204         if (tb[IFLA_LINKMODE])
3205                 dev->link_mode = nla_get_u8(tb[IFLA_LINKMODE]);
3206         if (tb[IFLA_GROUP])
3207                 dev_set_group(dev, nla_get_u32(tb[IFLA_GROUP]));
3208         if (tb[IFLA_GSO_MAX_SIZE])
3209                 netif_set_gso_max_size(dev, nla_get_u32(tb[IFLA_GSO_MAX_SIZE]));
3210         if (tb[IFLA_GSO_MAX_SEGS])
3211                 dev->gso_max_segs = nla_get_u32(tb[IFLA_GSO_MAX_SEGS]);
3212
3213         return dev;
3214 }
3215 EXPORT_SYMBOL(rtnl_create_link);
3216
3217 static int rtnl_group_changelink(const struct sk_buff *skb,
3218                 struct net *net, int group,
3219                 struct ifinfomsg *ifm,
3220                 struct netlink_ext_ack *extack,
3221                 struct nlattr **tb)
3222 {
3223         struct net_device *dev, *aux;
3224         int err;
3225
3226         for_each_netdev_safe(net, dev, aux) {
3227                 if (dev->group == group) {
3228                         err = do_setlink(skb, dev, ifm, extack, tb, NULL, 0);
3229                         if (err < 0)
3230                                 return err;
3231                 }
3232         }
3233
3234         return 0;
3235 }
3236
3237 static int __rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3238                           struct nlattr **attr, struct netlink_ext_ack *extack)
3239 {
3240         struct nlattr *slave_attr[RTNL_SLAVE_MAX_TYPE + 1];
3241         unsigned char name_assign_type = NET_NAME_USER;
3242         struct nlattr *linkinfo[IFLA_INFO_MAX + 1];
3243         const struct rtnl_link_ops *m_ops;
3244         struct net_device *master_dev;
3245         struct net *net = sock_net(skb->sk);
3246         const struct rtnl_link_ops *ops;
3247         struct nlattr *tb[IFLA_MAX + 1];
3248         struct net *dest_net, *link_net;
3249         struct nlattr **slave_data;
3250         char kind[MODULE_NAME_LEN];
3251         struct net_device *dev;
3252         struct ifinfomsg *ifm;
3253         char ifname[IFNAMSIZ];
3254         struct nlattr **data;
3255         int err;
3256
3257 #ifdef CONFIG_MODULES
3258 replay:
3259 #endif
3260         err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3261                                      ifla_policy, extack);
3262         if (err < 0)
3263                 return err;
3264
3265         err = rtnl_ensure_unique_netns(tb, extack, false);
3266         if (err < 0)
3267                 return err;
3268
3269         if (tb[IFLA_IFNAME])
3270                 nla_strlcpy(ifname, tb[IFLA_IFNAME], IFNAMSIZ);
3271         else
3272                 ifname[0] = '\0';
3273
3274         ifm = nlmsg_data(nlh);
3275         if (ifm->ifi_index > 0)
3276                 dev = __dev_get_by_index(net, ifm->ifi_index);
3277         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3278                 dev = rtnl_dev_get(net, NULL, tb[IFLA_ALT_IFNAME], ifname);
3279         else
3280                 dev = NULL;
3281
3282         master_dev = NULL;
3283         m_ops = NULL;
3284         if (dev) {
3285                 master_dev = netdev_master_upper_dev_get(dev);
3286                 if (master_dev)
3287                         m_ops = master_dev->rtnl_link_ops;
3288         }
3289
3290         err = validate_linkmsg(dev, tb);
3291         if (err < 0)
3292                 return err;
3293
3294         if (tb[IFLA_LINKINFO]) {
3295                 err = nla_parse_nested_deprecated(linkinfo, IFLA_INFO_MAX,
3296                                                   tb[IFLA_LINKINFO],
3297                                                   ifla_info_policy, NULL);
3298                 if (err < 0)
3299                         return err;
3300         } else
3301                 memset(linkinfo, 0, sizeof(linkinfo));
3302
3303         if (linkinfo[IFLA_INFO_KIND]) {
3304                 nla_strlcpy(kind, linkinfo[IFLA_INFO_KIND], sizeof(kind));
3305                 ops = rtnl_link_ops_get(kind);
3306         } else {
3307                 kind[0] = '\0';
3308                 ops = NULL;
3309         }
3310
3311         data = NULL;
3312         if (ops) {
3313                 if (ops->maxtype > RTNL_MAX_TYPE)
3314                         return -EINVAL;
3315
3316                 if (ops->maxtype && linkinfo[IFLA_INFO_DATA]) {
3317                         err = nla_parse_nested_deprecated(attr, ops->maxtype,
3318                                                           linkinfo[IFLA_INFO_DATA],
3319                                                           ops->policy, extack);
3320                         if (err < 0)
3321                                 return err;
3322                         data = attr;
3323                 }
3324                 if (ops->validate) {
3325                         err = ops->validate(tb, data, extack);
3326                         if (err < 0)
3327                                 return err;
3328                 }
3329         }
3330
3331         slave_data = NULL;
3332         if (m_ops) {
3333                 if (m_ops->slave_maxtype > RTNL_SLAVE_MAX_TYPE)
3334                         return -EINVAL;
3335
3336                 if (m_ops->slave_maxtype &&
3337                     linkinfo[IFLA_INFO_SLAVE_DATA]) {
3338                         err = nla_parse_nested_deprecated(slave_attr,
3339                                                           m_ops->slave_maxtype,
3340                                                           linkinfo[IFLA_INFO_SLAVE_DATA],
3341                                                           m_ops->slave_policy,
3342                                                           extack);
3343                         if (err < 0)
3344                                 return err;
3345                         slave_data = slave_attr;
3346                 }
3347         }
3348
3349         if (dev) {
3350                 int status = 0;
3351
3352                 if (nlh->nlmsg_flags & NLM_F_EXCL)
3353                         return -EEXIST;
3354                 if (nlh->nlmsg_flags & NLM_F_REPLACE)
3355                         return -EOPNOTSUPP;
3356
3357                 if (linkinfo[IFLA_INFO_DATA]) {
3358                         if (!ops || ops != dev->rtnl_link_ops ||
3359                             !ops->changelink)
3360                                 return -EOPNOTSUPP;
3361
3362                         err = ops->changelink(dev, tb, data, extack);
3363                         if (err < 0)
3364                                 return err;
3365                         status |= DO_SETLINK_NOTIFY;
3366                 }
3367
3368                 if (linkinfo[IFLA_INFO_SLAVE_DATA]) {
3369                         if (!m_ops || !m_ops->slave_changelink)
3370                                 return -EOPNOTSUPP;
3371
3372                         err = m_ops->slave_changelink(master_dev, dev, tb,
3373                                                       slave_data, extack);
3374                         if (err < 0)
3375                                 return err;
3376                         status |= DO_SETLINK_NOTIFY;
3377                 }
3378
3379                 return do_setlink(skb, dev, ifm, extack, tb, ifname, status);
3380         }
3381
3382         if (!(nlh->nlmsg_flags & NLM_F_CREATE)) {
3383                 if (ifm->ifi_index == 0 && tb[IFLA_GROUP])
3384                         return rtnl_group_changelink(skb, net,
3385                                                 nla_get_u32(tb[IFLA_GROUP]),
3386                                                 ifm, extack, tb);
3387                 return -ENODEV;
3388         }
3389
3390         if (tb[IFLA_MAP] || tb[IFLA_PROTINFO])
3391                 return -EOPNOTSUPP;
3392
3393         if (!ops) {
3394 #ifdef CONFIG_MODULES
3395                 if (kind[0]) {
3396                         __rtnl_unlock();
3397                         request_module("rtnl-link-%s", kind);
3398                         rtnl_lock();
3399                         ops = rtnl_link_ops_get(kind);
3400                         if (ops)
3401                                 goto replay;
3402                 }
3403 #endif
3404                 NL_SET_ERR_MSG(extack, "Unknown device type");
3405                 return -EOPNOTSUPP;
3406         }
3407
3408         if (!ops->setup)
3409                 return -EOPNOTSUPP;
3410
3411         if (!ifname[0]) {
3412                 snprintf(ifname, IFNAMSIZ, "%s%%d", ops->kind);
3413                 name_assign_type = NET_NAME_ENUM;
3414         }
3415
3416         dest_net = rtnl_link_get_net_capable(skb, net, tb, CAP_NET_ADMIN);
3417         if (IS_ERR(dest_net))
3418                 return PTR_ERR(dest_net);
3419
3420         if (tb[IFLA_LINK_NETNSID]) {
3421                 int id = nla_get_s32(tb[IFLA_LINK_NETNSID]);
3422
3423                 link_net = get_net_ns_by_id(dest_net, id);
3424                 if (!link_net) {
3425                         NL_SET_ERR_MSG(extack, "Unknown network namespace id");
3426                         err =  -EINVAL;
3427                         goto out;
3428                 }
3429                 err = -EPERM;
3430                 if (!netlink_ns_capable(skb, link_net->user_ns, CAP_NET_ADMIN))
3431                         goto out;
3432         } else {
3433                 link_net = NULL;
3434         }
3435
3436         dev = rtnl_create_link(link_net ? : dest_net, ifname,
3437                                name_assign_type, ops, tb, extack);
3438         if (IS_ERR(dev)) {
3439                 err = PTR_ERR(dev);
3440                 goto out;
3441         }
3442
3443         dev->ifindex = ifm->ifi_index;
3444
3445         if (ops->newlink)
3446                 err = ops->newlink(link_net ? : net, dev, tb, data, extack);
3447         else
3448                 err = register_netdevice(dev);
3449         if (err < 0) {
3450                 free_netdev(dev);
3451                 goto out;
3452         }
3453
3454         err = rtnl_configure_link(dev, ifm);
3455         if (err < 0)
3456                 goto out_unregister;
3457         if (link_net) {
3458                 err = dev_change_net_namespace(dev, dest_net, ifname);
3459                 if (err < 0)
3460                         goto out_unregister;
3461         }
3462         if (tb[IFLA_MASTER]) {
3463                 err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
3464                 if (err)
3465                         goto out_unregister;
3466         }
3467 out:
3468         if (link_net)
3469                 put_net(link_net);
3470         put_net(dest_net);
3471         return err;
3472 out_unregister:
3473         if (ops->newlink) {
3474                 LIST_HEAD(list_kill);
3475
3476                 ops->dellink(dev, &list_kill);
3477                 unregister_netdevice_many(&list_kill);
3478         } else {
3479                 unregister_netdevice(dev);
3480         }
3481         goto out;
3482 }
3483
3484 static int rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3485                         struct netlink_ext_ack *extack)
3486 {
3487         struct nlattr **attr;
3488         int ret;
3489
3490         attr = kmalloc_array(RTNL_MAX_TYPE + 1, sizeof(*attr), GFP_KERNEL);
3491         if (!attr)
3492                 return -ENOMEM;
3493
3494         ret = __rtnl_newlink(skb, nlh, attr, extack);
3495         kfree(attr);
3496         return ret;
3497 }
3498
3499 static int rtnl_valid_getlink_req(struct sk_buff *skb,
3500                                   const struct nlmsghdr *nlh,
3501                                   struct nlattr **tb,
3502                                   struct netlink_ext_ack *extack)
3503 {
3504         struct ifinfomsg *ifm;
3505         int i, err;
3506
3507         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
3508                 NL_SET_ERR_MSG(extack, "Invalid header for get link");
3509                 return -EINVAL;
3510         }
3511
3512         if (!netlink_strict_get_check(skb))
3513                 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFLA_MAX,
3514                                               ifla_policy, extack);
3515
3516         ifm = nlmsg_data(nlh);
3517         if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
3518             ifm->ifi_change) {
3519                 NL_SET_ERR_MSG(extack, "Invalid values in header for get link request");
3520                 return -EINVAL;
3521         }
3522
3523         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFLA_MAX,
3524                                             ifla_policy, extack);
3525         if (err)
3526                 return err;
3527
3528         for (i = 0; i <= IFLA_MAX; i++) {
3529                 if (!tb[i])
3530                         continue;
3531
3532                 switch (i) {
3533                 case IFLA_IFNAME:
3534                 case IFLA_ALT_IFNAME:
3535                 case IFLA_EXT_MASK:
3536                 case IFLA_TARGET_NETNSID:
3537                         break;
3538                 default:
3539                         NL_SET_ERR_MSG(extack, "Unsupported attribute in get link request");
3540                         return -EINVAL;
3541                 }
3542         }
3543
3544         return 0;
3545 }
3546
3547 static int rtnl_getlink(struct sk_buff *skb, struct nlmsghdr *nlh,
3548                         struct netlink_ext_ack *extack)
3549 {
3550         struct net *net = sock_net(skb->sk);
3551         struct net *tgt_net = net;
3552         struct ifinfomsg *ifm;
3553         struct nlattr *tb[IFLA_MAX+1];
3554         struct net_device *dev = NULL;
3555         struct sk_buff *nskb;
3556         int netnsid = -1;
3557         int err;
3558         u32 ext_filter_mask = 0;
3559
3560         err = rtnl_valid_getlink_req(skb, nlh, tb, extack);
3561         if (err < 0)
3562                 return err;
3563
3564         err = rtnl_ensure_unique_netns(tb, extack, true);
3565         if (err < 0)
3566                 return err;
3567
3568         if (tb[IFLA_TARGET_NETNSID]) {
3569                 netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
3570                 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
3571                 if (IS_ERR(tgt_net))
3572                         return PTR_ERR(tgt_net);
3573         }
3574
3575         if (tb[IFLA_EXT_MASK])
3576                 ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3577
3578         err = -EINVAL;
3579         ifm = nlmsg_data(nlh);
3580         if (ifm->ifi_index > 0)
3581                 dev = __dev_get_by_index(tgt_net, ifm->ifi_index);
3582         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3583                 dev = rtnl_dev_get(tgt_net, tb[IFLA_IFNAME],
3584                                    tb[IFLA_ALT_IFNAME], NULL);
3585         else
3586                 goto out;
3587
3588         err = -ENODEV;
3589         if (dev == NULL)
3590                 goto out;
3591
3592         err = -ENOBUFS;
3593         nskb = nlmsg_new(if_nlmsg_size(dev, ext_filter_mask), GFP_KERNEL);
3594         if (nskb == NULL)
3595                 goto out;
3596
3597         err = rtnl_fill_ifinfo(nskb, dev, net,
3598                                RTM_NEWLINK, NETLINK_CB(skb).portid,
3599                                nlh->nlmsg_seq, 0, 0, ext_filter_mask,
3600                                0, NULL, 0, netnsid, GFP_KERNEL);
3601         if (err < 0) {
3602                 /* -EMSGSIZE implies BUG in if_nlmsg_size */
3603                 WARN_ON(err == -EMSGSIZE);
3604                 kfree_skb(nskb);
3605         } else
3606                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
3607 out:
3608         if (netnsid >= 0)
3609                 put_net(tgt_net);
3610
3611         return err;
3612 }
3613
3614 static int rtnl_alt_ifname(int cmd, struct net_device *dev, struct nlattr *attr,
3615                            bool *changed, struct netlink_ext_ack *extack)
3616 {
3617         char *alt_ifname;
3618         size_t size;
3619         int err;
3620
3621         err = nla_validate(attr, attr->nla_len, IFLA_MAX, ifla_policy, extack);
3622         if (err)
3623                 return err;
3624
3625         if (cmd == RTM_NEWLINKPROP) {
3626                 size = rtnl_prop_list_size(dev);
3627                 size += nla_total_size(ALTIFNAMSIZ);
3628                 if (size >= U16_MAX) {
3629                         NL_SET_ERR_MSG(extack,
3630                                        "effective property list too long");
3631                         return -EINVAL;
3632                 }
3633         }
3634
3635         alt_ifname = nla_strdup(attr, GFP_KERNEL_ACCOUNT);
3636         if (!alt_ifname)
3637                 return -ENOMEM;
3638
3639         if (cmd == RTM_NEWLINKPROP) {
3640                 err = netdev_name_node_alt_create(dev, alt_ifname);
3641                 if (!err)
3642                         alt_ifname = NULL;
3643         } else if (cmd == RTM_DELLINKPROP) {
3644                 err = netdev_name_node_alt_destroy(dev, alt_ifname);
3645         } else {
3646                 WARN_ON_ONCE(1);
3647                 err = -EINVAL;
3648         }
3649
3650         kfree(alt_ifname);
3651         if (!err)
3652                 *changed = true;
3653         return err;
3654 }
3655
3656 static int rtnl_linkprop(int cmd, struct sk_buff *skb, struct nlmsghdr *nlh,
3657                          struct netlink_ext_ack *extack)
3658 {
3659         struct net *net = sock_net(skb->sk);
3660         struct nlattr *tb[IFLA_MAX + 1];
3661         struct net_device *dev;
3662         struct ifinfomsg *ifm;
3663         bool changed = false;
3664         struct nlattr *attr;
3665         int err, rem;
3666
3667         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFLA_MAX, ifla_policy, extack);
3668         if (err)
3669                 return err;
3670
3671         err = rtnl_ensure_unique_netns(tb, extack, true);
3672         if (err)
3673                 return err;
3674
3675         ifm = nlmsg_data(nlh);
3676         if (ifm->ifi_index > 0)
3677                 dev = __dev_get_by_index(net, ifm->ifi_index);
3678         else if (tb[IFLA_IFNAME] || tb[IFLA_ALT_IFNAME])
3679                 dev = rtnl_dev_get(net, tb[IFLA_IFNAME],
3680                                    tb[IFLA_ALT_IFNAME], NULL);
3681         else
3682                 return -EINVAL;
3683
3684         if (!dev)
3685                 return -ENODEV;
3686
3687         if (!tb[IFLA_PROP_LIST])
3688                 return 0;
3689
3690         nla_for_each_nested(attr, tb[IFLA_PROP_LIST], rem) {
3691                 switch (nla_type(attr)) {
3692                 case IFLA_ALT_IFNAME:
3693                         err = rtnl_alt_ifname(cmd, dev, attr, &changed, extack);
3694                         if (err)
3695                                 return err;
3696                         break;
3697                 }
3698         }
3699
3700         if (changed)
3701                 netdev_state_change(dev);
3702         return 0;
3703 }
3704
3705 static int rtnl_newlinkprop(struct sk_buff *skb, struct nlmsghdr *nlh,
3706                             struct netlink_ext_ack *extack)
3707 {
3708         return rtnl_linkprop(RTM_NEWLINKPROP, skb, nlh, extack);
3709 }
3710
3711 static int rtnl_dellinkprop(struct sk_buff *skb, struct nlmsghdr *nlh,
3712                             struct netlink_ext_ack *extack)
3713 {
3714         return rtnl_linkprop(RTM_DELLINKPROP, skb, nlh, extack);
3715 }
3716
3717 static u32 rtnl_calcit(struct sk_buff *skb, struct nlmsghdr *nlh)
3718 {
3719         struct net *net = sock_net(skb->sk);
3720         size_t min_ifinfo_dump_size = 0;
3721         struct nlattr *tb[IFLA_MAX+1];
3722         u32 ext_filter_mask = 0;
3723         struct net_device *dev;
3724         int hdrlen;
3725
3726         /* Same kernel<->userspace interface hack as in rtnl_dump_ifinfo. */
3727         hdrlen = nlmsg_len(nlh) < sizeof(struct ifinfomsg) ?
3728                  sizeof(struct rtgenmsg) : sizeof(struct ifinfomsg);
3729
3730         if (nlmsg_parse_deprecated(nlh, hdrlen, tb, IFLA_MAX, ifla_policy, NULL) >= 0) {
3731                 if (tb[IFLA_EXT_MASK])
3732                         ext_filter_mask = nla_get_u32(tb[IFLA_EXT_MASK]);
3733         }
3734
3735         if (!ext_filter_mask)
3736                 return NLMSG_GOODSIZE;
3737         /*
3738          * traverse the list of net devices and compute the minimum
3739          * buffer size based upon the filter mask.
3740          */
3741         rcu_read_lock();
3742         for_each_netdev_rcu(net, dev) {
3743                 min_ifinfo_dump_size = max(min_ifinfo_dump_size,
3744                                            if_nlmsg_size(dev, ext_filter_mask));
3745         }
3746         rcu_read_unlock();
3747
3748         return nlmsg_total_size(min_ifinfo_dump_size);
3749 }
3750
3751 static int rtnl_dump_all(struct sk_buff *skb, struct netlink_callback *cb)
3752 {
3753         int idx;
3754         int s_idx = cb->family;
3755         int type = cb->nlh->nlmsg_type - RTM_BASE;
3756         int ret = 0;
3757
3758         if (s_idx == 0)
3759                 s_idx = 1;
3760
3761         for (idx = 1; idx <= RTNL_FAMILY_MAX; idx++) {
3762                 struct rtnl_link **tab;
3763                 struct rtnl_link *link;
3764                 rtnl_dumpit_func dumpit;
3765
3766                 if (idx < s_idx || idx == PF_PACKET)
3767                         continue;
3768
3769                 if (type < 0 || type >= RTM_NR_MSGTYPES)
3770                         continue;
3771
3772                 tab = rcu_dereference_rtnl(rtnl_msg_handlers[idx]);
3773                 if (!tab)
3774                         continue;
3775
3776                 link = tab[type];
3777                 if (!link)
3778                         continue;
3779
3780                 dumpit = link->dumpit;
3781                 if (!dumpit)
3782                         continue;
3783
3784                 if (idx > s_idx) {
3785                         memset(&cb->args[0], 0, sizeof(cb->args));
3786                         cb->prev_seq = 0;
3787                         cb->seq = 0;
3788                 }
3789                 ret = dumpit(skb, cb);
3790                 if (ret)
3791                         break;
3792         }
3793         cb->family = idx;
3794
3795         return skb->len ? : ret;
3796 }
3797
3798 struct sk_buff *rtmsg_ifinfo_build_skb(int type, struct net_device *dev,
3799                                        unsigned int change,
3800                                        u32 event, gfp_t flags, int *new_nsid,
3801                                        int new_ifindex)
3802 {
3803         struct net *net = dev_net(dev);
3804         struct sk_buff *skb;
3805         int err = -ENOBUFS;
3806         size_t if_info_size;
3807
3808         skb = nlmsg_new((if_info_size = if_nlmsg_size(dev, 0)), flags);
3809         if (skb == NULL)
3810                 goto errout;
3811
3812         err = rtnl_fill_ifinfo(skb, dev, dev_net(dev),
3813                                type, 0, 0, change, 0, 0, event,
3814                                new_nsid, new_ifindex, -1, flags);
3815         if (err < 0) {
3816                 /* -EMSGSIZE implies BUG in if_nlmsg_size() */
3817                 WARN_ON(err == -EMSGSIZE);
3818                 kfree_skb(skb);
3819                 goto errout;
3820         }
3821         return skb;
3822 errout:
3823         if (err < 0)
3824                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
3825         return NULL;
3826 }
3827
3828 void rtmsg_ifinfo_send(struct sk_buff *skb, struct net_device *dev, gfp_t flags)
3829 {
3830         struct net *net = dev_net(dev);
3831
3832         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, flags);
3833 }
3834
3835 static void rtmsg_ifinfo_event(int type, struct net_device *dev,
3836                                unsigned int change, u32 event,
3837                                gfp_t flags, int *new_nsid, int new_ifindex)
3838 {
3839         struct sk_buff *skb;
3840
3841         if (dev->reg_state != NETREG_REGISTERED)
3842                 return;
3843
3844         skb = rtmsg_ifinfo_build_skb(type, dev, change, event, flags, new_nsid,
3845                                      new_ifindex);
3846         if (skb)
3847                 rtmsg_ifinfo_send(skb, dev, flags);
3848 }
3849
3850 void rtmsg_ifinfo(int type, struct net_device *dev, unsigned int change,
3851                   gfp_t flags)
3852 {
3853         rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
3854                            NULL, 0);
3855 }
3856
3857 void rtmsg_ifinfo_newnet(int type, struct net_device *dev, unsigned int change,
3858                          gfp_t flags, int *new_nsid, int new_ifindex)
3859 {
3860         rtmsg_ifinfo_event(type, dev, change, rtnl_get_event(0), flags,
3861                            new_nsid, new_ifindex);
3862 }
3863
3864 static int nlmsg_populate_fdb_fill(struct sk_buff *skb,
3865                                    struct net_device *dev,
3866                                    u8 *addr, u16 vid, u32 pid, u32 seq,
3867                                    int type, unsigned int flags,
3868                                    int nlflags, u16 ndm_state)
3869 {
3870         struct nlmsghdr *nlh;
3871         struct ndmsg *ndm;
3872
3873         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ndm), nlflags);
3874         if (!nlh)
3875                 return -EMSGSIZE;
3876
3877         ndm = nlmsg_data(nlh);
3878         ndm->ndm_family  = AF_BRIDGE;
3879         ndm->ndm_pad1    = 0;
3880         ndm->ndm_pad2    = 0;
3881         ndm->ndm_flags   = flags;
3882         ndm->ndm_type    = 0;
3883         ndm->ndm_ifindex = dev->ifindex;
3884         ndm->ndm_state   = ndm_state;
3885
3886         if (nla_put(skb, NDA_LLADDR, ETH_ALEN, addr))
3887                 goto nla_put_failure;
3888         if (vid)
3889                 if (nla_put(skb, NDA_VLAN, sizeof(u16), &vid))
3890                         goto nla_put_failure;
3891
3892         nlmsg_end(skb, nlh);
3893         return 0;
3894
3895 nla_put_failure:
3896         nlmsg_cancel(skb, nlh);
3897         return -EMSGSIZE;
3898 }
3899
3900 static inline size_t rtnl_fdb_nlmsg_size(void)
3901 {
3902         return NLMSG_ALIGN(sizeof(struct ndmsg)) +
3903                nla_total_size(ETH_ALEN) +       /* NDA_LLADDR */
3904                nla_total_size(sizeof(u16)) +    /* NDA_VLAN */
3905                0;
3906 }
3907
3908 static void rtnl_fdb_notify(struct net_device *dev, u8 *addr, u16 vid, int type,
3909                             u16 ndm_state)
3910 {
3911         struct net *net = dev_net(dev);
3912         struct sk_buff *skb;
3913         int err = -ENOBUFS;
3914
3915         skb = nlmsg_new(rtnl_fdb_nlmsg_size(), GFP_ATOMIC);
3916         if (!skb)
3917                 goto errout;
3918
3919         err = nlmsg_populate_fdb_fill(skb, dev, addr, vid,
3920                                       0, 0, type, NTF_SELF, 0, ndm_state);
3921         if (err < 0) {
3922                 kfree_skb(skb);
3923                 goto errout;
3924         }
3925
3926         rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
3927         return;
3928 errout:
3929         rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
3930 }
3931
3932 /*
3933  * ndo_dflt_fdb_add - default netdevice operation to add an FDB entry
3934  */
3935 int ndo_dflt_fdb_add(struct ndmsg *ndm,
3936                      struct nlattr *tb[],
3937                      struct net_device *dev,
3938                      const unsigned char *addr, u16 vid,
3939                      u16 flags)
3940 {
3941         int err = -EINVAL;
3942
3943         /* If aging addresses are supported device will need to
3944          * implement its own handler for this.
3945          */
3946         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
3947                 pr_info("%s: FDB only supports static addresses\n", dev->name);
3948                 return err;
3949         }
3950
3951         if (vid) {
3952                 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
3953                 return err;
3954         }
3955
3956         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
3957                 err = dev_uc_add_excl(dev, addr);
3958         else if (is_multicast_ether_addr(addr))
3959                 err = dev_mc_add_excl(dev, addr);
3960
3961         /* Only return duplicate errors if NLM_F_EXCL is set */
3962         if (err == -EEXIST && !(flags & NLM_F_EXCL))
3963                 err = 0;
3964
3965         return err;
3966 }
3967 EXPORT_SYMBOL(ndo_dflt_fdb_add);
3968
3969 static int fdb_vid_parse(struct nlattr *vlan_attr, u16 *p_vid,
3970                          struct netlink_ext_ack *extack)
3971 {
3972         u16 vid = 0;
3973
3974         if (vlan_attr) {
3975                 if (nla_len(vlan_attr) != sizeof(u16)) {
3976                         NL_SET_ERR_MSG(extack, "invalid vlan attribute size");
3977                         return -EINVAL;
3978                 }
3979
3980                 vid = nla_get_u16(vlan_attr);
3981
3982                 if (!vid || vid >= VLAN_VID_MASK) {
3983                         NL_SET_ERR_MSG(extack, "invalid vlan id");
3984                         return -EINVAL;
3985                 }
3986         }
3987         *p_vid = vid;
3988         return 0;
3989 }
3990
3991 static int rtnl_fdb_add(struct sk_buff *skb, struct nlmsghdr *nlh,
3992                         struct netlink_ext_ack *extack)
3993 {
3994         struct net *net = sock_net(skb->sk);
3995         struct ndmsg *ndm;
3996         struct nlattr *tb[NDA_MAX+1];
3997         struct net_device *dev;
3998         u8 *addr;
3999         u16 vid;
4000         int err;
4001
4002         err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX, NULL,
4003                                      extack);
4004         if (err < 0)
4005                 return err;
4006
4007         ndm = nlmsg_data(nlh);
4008         if (ndm->ndm_ifindex == 0) {
4009                 NL_SET_ERR_MSG(extack, "invalid ifindex");
4010                 return -EINVAL;
4011         }
4012
4013         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
4014         if (dev == NULL) {
4015                 NL_SET_ERR_MSG(extack, "unknown ifindex");
4016                 return -ENODEV;
4017         }
4018
4019         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
4020                 NL_SET_ERR_MSG(extack, "invalid address");
4021                 return -EINVAL;
4022         }
4023
4024         if (dev->type != ARPHRD_ETHER) {
4025                 NL_SET_ERR_MSG(extack, "FDB add only supported for Ethernet devices");
4026                 return -EINVAL;
4027         }
4028
4029         addr = nla_data(tb[NDA_LLADDR]);
4030
4031         err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
4032         if (err)
4033                 return err;
4034
4035         err = -EOPNOTSUPP;
4036
4037         /* Support fdb on master device the net/bridge default case */
4038         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
4039             netif_is_bridge_port(dev)) {
4040                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4041                 const struct net_device_ops *ops = br_dev->netdev_ops;
4042
4043                 err = ops->ndo_fdb_add(ndm, tb, dev, addr, vid,
4044                                        nlh->nlmsg_flags, extack);
4045                 if (err)
4046                         goto out;
4047                 else
4048                         ndm->ndm_flags &= ~NTF_MASTER;
4049         }
4050
4051         /* Embedded bridge, macvlan, and any other device support */
4052         if ((ndm->ndm_flags & NTF_SELF)) {
4053                 if (dev->netdev_ops->ndo_fdb_add)
4054                         err = dev->netdev_ops->ndo_fdb_add(ndm, tb, dev, addr,
4055                                                            vid,
4056                                                            nlh->nlmsg_flags,
4057                                                            extack);
4058                 else
4059                         err = ndo_dflt_fdb_add(ndm, tb, dev, addr, vid,
4060                                                nlh->nlmsg_flags);
4061
4062                 if (!err) {
4063                         rtnl_fdb_notify(dev, addr, vid, RTM_NEWNEIGH,
4064                                         ndm->ndm_state);
4065                         ndm->ndm_flags &= ~NTF_SELF;
4066                 }
4067         }
4068 out:
4069         return err;
4070 }
4071
4072 /*
4073  * ndo_dflt_fdb_del - default netdevice operation to delete an FDB entry
4074  */
4075 int ndo_dflt_fdb_del(struct ndmsg *ndm,
4076                      struct nlattr *tb[],
4077                      struct net_device *dev,
4078                      const unsigned char *addr, u16 vid)
4079 {
4080         int err = -EINVAL;
4081
4082         /* If aging addresses are supported device will need to
4083          * implement its own handler for this.
4084          */
4085         if (!(ndm->ndm_state & NUD_PERMANENT)) {
4086                 pr_info("%s: FDB only supports static addresses\n", dev->name);
4087                 return err;
4088         }
4089
4090         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
4091                 err = dev_uc_del(dev, addr);
4092         else if (is_multicast_ether_addr(addr))
4093                 err = dev_mc_del(dev, addr);
4094
4095         return err;
4096 }
4097 EXPORT_SYMBOL(ndo_dflt_fdb_del);
4098
4099 static int rtnl_fdb_del(struct sk_buff *skb, struct nlmsghdr *nlh,
4100                         struct netlink_ext_ack *extack)
4101 {
4102         struct net *net = sock_net(skb->sk);
4103         struct ndmsg *ndm;
4104         struct nlattr *tb[NDA_MAX+1];
4105         struct net_device *dev;
4106         __u8 *addr;
4107         int err;
4108         u16 vid;
4109
4110         if (!netlink_capable(skb, CAP_NET_ADMIN))
4111                 return -EPERM;
4112
4113         err = nlmsg_parse_deprecated(nlh, sizeof(*ndm), tb, NDA_MAX, NULL,
4114                                      extack);
4115         if (err < 0)
4116                 return err;
4117
4118         ndm = nlmsg_data(nlh);
4119         if (ndm->ndm_ifindex == 0) {
4120                 NL_SET_ERR_MSG(extack, "invalid ifindex");
4121                 return -EINVAL;
4122         }
4123
4124         dev = __dev_get_by_index(net, ndm->ndm_ifindex);
4125         if (dev == NULL) {
4126                 NL_SET_ERR_MSG(extack, "unknown ifindex");
4127                 return -ENODEV;
4128         }
4129
4130         if (!tb[NDA_LLADDR] || nla_len(tb[NDA_LLADDR]) != ETH_ALEN) {
4131                 NL_SET_ERR_MSG(extack, "invalid address");
4132                 return -EINVAL;
4133         }
4134
4135         if (dev->type != ARPHRD_ETHER) {
4136                 NL_SET_ERR_MSG(extack, "FDB delete only supported for Ethernet devices");
4137                 return -EINVAL;
4138         }
4139
4140         addr = nla_data(tb[NDA_LLADDR]);
4141
4142         err = fdb_vid_parse(tb[NDA_VLAN], &vid, extack);
4143         if (err)
4144                 return err;
4145
4146         err = -EOPNOTSUPP;
4147
4148         /* Support fdb on master device the net/bridge default case */
4149         if ((!ndm->ndm_flags || ndm->ndm_flags & NTF_MASTER) &&
4150             netif_is_bridge_port(dev)) {
4151                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4152                 const struct net_device_ops *ops = br_dev->netdev_ops;
4153
4154                 if (ops->ndo_fdb_del)
4155                         err = ops->ndo_fdb_del(ndm, tb, dev, addr, vid);
4156
4157                 if (err)
4158                         goto out;
4159                 else
4160                         ndm->ndm_flags &= ~NTF_MASTER;
4161         }
4162
4163         /* Embedded bridge, macvlan, and any other device support */
4164         if (ndm->ndm_flags & NTF_SELF) {
4165                 if (dev->netdev_ops->ndo_fdb_del)
4166                         err = dev->netdev_ops->ndo_fdb_del(ndm, tb, dev, addr,
4167                                                            vid);
4168                 else
4169                         err = ndo_dflt_fdb_del(ndm, tb, dev, addr, vid);
4170
4171                 if (!err) {
4172                         rtnl_fdb_notify(dev, addr, vid, RTM_DELNEIGH,
4173                                         ndm->ndm_state);
4174                         ndm->ndm_flags &= ~NTF_SELF;
4175                 }
4176         }
4177 out:
4178         return err;
4179 }
4180
4181 static int nlmsg_populate_fdb(struct sk_buff *skb,
4182                               struct netlink_callback *cb,
4183                               struct net_device *dev,
4184                               int *idx,
4185                               struct netdev_hw_addr_list *list)
4186 {
4187         struct netdev_hw_addr *ha;
4188         int err;
4189         u32 portid, seq;
4190
4191         portid = NETLINK_CB(cb->skb).portid;
4192         seq = cb->nlh->nlmsg_seq;
4193
4194         list_for_each_entry(ha, &list->list, list) {
4195                 if (*idx < cb->args[2])
4196                         goto skip;
4197
4198                 err = nlmsg_populate_fdb_fill(skb, dev, ha->addr, 0,
4199                                               portid, seq,
4200                                               RTM_NEWNEIGH, NTF_SELF,
4201                                               NLM_F_MULTI, NUD_PERMANENT);
4202                 if (err < 0)
4203                         return err;
4204 skip:
4205                 *idx += 1;
4206         }
4207         return 0;
4208 }
4209
4210 /**
4211  * ndo_dflt_fdb_dump - default netdevice operation to dump an FDB table.
4212  * @skb: socket buffer to store message in
4213  * @cb: netlink callback
4214  * @dev: netdevice
4215  * @filter_dev: ignored
4216  * @idx: the number of FDB table entries dumped is added to *@idx
4217  *
4218  * Default netdevice operation to dump the existing unicast address list.
4219  * Returns number of addresses from list put in skb.
4220  */
4221 int ndo_dflt_fdb_dump(struct sk_buff *skb,
4222                       struct netlink_callback *cb,
4223                       struct net_device *dev,
4224                       struct net_device *filter_dev,
4225                       int *idx)
4226 {
4227         int err;
4228
4229         if (dev->type != ARPHRD_ETHER)
4230                 return -EINVAL;
4231
4232         netif_addr_lock_bh(dev);
4233         err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->uc);
4234         if (err)
4235                 goto out;
4236         err = nlmsg_populate_fdb(skb, cb, dev, idx, &dev->mc);
4237 out:
4238         netif_addr_unlock_bh(dev);
4239         return err;
4240 }
4241 EXPORT_SYMBOL(ndo_dflt_fdb_dump);
4242
4243 static int valid_fdb_dump_strict(const struct nlmsghdr *nlh,
4244                                  int *br_idx, int *brport_idx,
4245                                  struct netlink_ext_ack *extack)
4246 {
4247         struct nlattr *tb[NDA_MAX + 1];
4248         struct ndmsg *ndm;
4249         int err, i;
4250
4251         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
4252                 NL_SET_ERR_MSG(extack, "Invalid header for fdb dump request");
4253                 return -EINVAL;
4254         }
4255
4256         ndm = nlmsg_data(nlh);
4257         if (ndm->ndm_pad1  || ndm->ndm_pad2  || ndm->ndm_state ||
4258             ndm->ndm_flags || ndm->ndm_type) {
4259                 NL_SET_ERR_MSG(extack, "Invalid values in header for fdb dump request");
4260                 return -EINVAL;
4261         }
4262
4263         err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
4264                                             NDA_MAX, NULL, extack);
4265         if (err < 0)
4266                 return err;
4267
4268         *brport_idx = ndm->ndm_ifindex;
4269         for (i = 0; i <= NDA_MAX; ++i) {
4270                 if (!tb[i])
4271                         continue;
4272
4273                 switch (i) {
4274                 case NDA_IFINDEX:
4275                         if (nla_len(tb[i]) != sizeof(u32)) {
4276                                 NL_SET_ERR_MSG(extack, "Invalid IFINDEX attribute in fdb dump request");
4277                                 return -EINVAL;
4278                         }
4279                         *brport_idx = nla_get_u32(tb[NDA_IFINDEX]);
4280                         break;
4281                 case NDA_MASTER:
4282                         if (nla_len(tb[i]) != sizeof(u32)) {
4283                                 NL_SET_ERR_MSG(extack, "Invalid MASTER attribute in fdb dump request");
4284                                 return -EINVAL;
4285                         }
4286                         *br_idx = nla_get_u32(tb[NDA_MASTER]);
4287                         break;
4288                 default:
4289                         NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb dump request");
4290                         return -EINVAL;
4291                 }
4292         }
4293
4294         return 0;
4295 }
4296
4297 static int valid_fdb_dump_legacy(const struct nlmsghdr *nlh,
4298                                  int *br_idx, int *brport_idx,
4299                                  struct netlink_ext_ack *extack)
4300 {
4301         struct nlattr *tb[IFLA_MAX+1];
4302         int err;
4303
4304         /* A hack to preserve kernel<->userspace interface.
4305          * Before Linux v4.12 this code accepted ndmsg since iproute2 v3.3.0.
4306          * However, ndmsg is shorter than ifinfomsg thus nlmsg_parse() bails.
4307          * So, check for ndmsg with an optional u32 attribute (not used here).
4308          * Fortunately these sizes don't conflict with the size of ifinfomsg
4309          * with an optional attribute.
4310          */
4311         if (nlmsg_len(nlh) != sizeof(struct ndmsg) &&
4312             (nlmsg_len(nlh) != sizeof(struct ndmsg) +
4313              nla_attr_size(sizeof(u32)))) {
4314                 struct ifinfomsg *ifm;
4315
4316                 err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg),
4317                                              tb, IFLA_MAX, ifla_policy,
4318                                              extack);
4319                 if (err < 0) {
4320                         return -EINVAL;
4321                 } else if (err == 0) {
4322                         if (tb[IFLA_MASTER])
4323                                 *br_idx = nla_get_u32(tb[IFLA_MASTER]);
4324                 }
4325
4326                 ifm = nlmsg_data(nlh);
4327                 *brport_idx = ifm->ifi_index;
4328         }
4329         return 0;
4330 }
4331
4332 static int rtnl_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb)
4333 {
4334         struct net_device *dev;
4335         struct net_device *br_dev = NULL;
4336         const struct net_device_ops *ops = NULL;
4337         const struct net_device_ops *cops = NULL;
4338         struct net *net = sock_net(skb->sk);
4339         struct hlist_head *head;
4340         int brport_idx = 0;
4341         int br_idx = 0;
4342         int h, s_h;
4343         int idx = 0, s_idx;
4344         int err = 0;
4345         int fidx = 0;
4346
4347         if (cb->strict_check)
4348                 err = valid_fdb_dump_strict(cb->nlh, &br_idx, &brport_idx,
4349                                             cb->extack);
4350         else
4351                 err = valid_fdb_dump_legacy(cb->nlh, &br_idx, &brport_idx,
4352                                             cb->extack);
4353         if (err < 0)
4354                 return err;
4355
4356         if (br_idx) {
4357                 br_dev = __dev_get_by_index(net, br_idx);
4358                 if (!br_dev)
4359                         return -ENODEV;
4360
4361                 ops = br_dev->netdev_ops;
4362         }
4363
4364         s_h = cb->args[0];
4365         s_idx = cb->args[1];
4366
4367         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4368                 idx = 0;
4369                 head = &net->dev_index_head[h];
4370                 hlist_for_each_entry(dev, head, index_hlist) {
4371
4372                         if (brport_idx && (dev->ifindex != brport_idx))
4373                                 continue;
4374
4375                         if (!br_idx) { /* user did not specify a specific bridge */
4376                                 if (netif_is_bridge_port(dev)) {
4377                                         br_dev = netdev_master_upper_dev_get(dev);
4378                                         cops = br_dev->netdev_ops;
4379                                 }
4380                         } else {
4381                                 if (dev != br_dev &&
4382                                     !netif_is_bridge_port(dev))
4383                                         continue;
4384
4385                                 if (br_dev != netdev_master_upper_dev_get(dev) &&
4386                                     !(dev->priv_flags & IFF_EBRIDGE))
4387                                         continue;
4388                                 cops = ops;
4389                         }
4390
4391                         if (idx < s_idx)
4392                                 goto cont;
4393
4394                         if (netif_is_bridge_port(dev)) {
4395                                 if (cops && cops->ndo_fdb_dump) {
4396                                         err = cops->ndo_fdb_dump(skb, cb,
4397                                                                 br_dev, dev,
4398                                                                 &fidx);
4399                                         if (err == -EMSGSIZE)
4400                                                 goto out;
4401                                 }
4402                         }
4403
4404                         if (dev->netdev_ops->ndo_fdb_dump)
4405                                 err = dev->netdev_ops->ndo_fdb_dump(skb, cb,
4406                                                                     dev, NULL,
4407                                                                     &fidx);
4408                         else
4409                                 err = ndo_dflt_fdb_dump(skb, cb, dev, NULL,
4410                                                         &fidx);
4411                         if (err == -EMSGSIZE)
4412                                 goto out;
4413
4414                         cops = NULL;
4415
4416                         /* reset fdb offset to 0 for rest of the interfaces */
4417                         cb->args[2] = 0;
4418                         fidx = 0;
4419 cont:
4420                         idx++;
4421                 }
4422         }
4423
4424 out:
4425         cb->args[0] = h;
4426         cb->args[1] = idx;
4427         cb->args[2] = fidx;
4428
4429         return skb->len;
4430 }
4431
4432 static int valid_fdb_get_strict(const struct nlmsghdr *nlh,
4433                                 struct nlattr **tb, u8 *ndm_flags,
4434                                 int *br_idx, int *brport_idx, u8 **addr,
4435                                 u16 *vid, struct netlink_ext_ack *extack)
4436 {
4437         struct ndmsg *ndm;
4438         int err, i;
4439
4440         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ndm))) {
4441                 NL_SET_ERR_MSG(extack, "Invalid header for fdb get request");
4442                 return -EINVAL;
4443         }
4444
4445         ndm = nlmsg_data(nlh);
4446         if (ndm->ndm_pad1  || ndm->ndm_pad2  || ndm->ndm_state ||
4447             ndm->ndm_type) {
4448                 NL_SET_ERR_MSG(extack, "Invalid values in header for fdb get request");
4449                 return -EINVAL;
4450         }
4451
4452         if (ndm->ndm_flags & ~(NTF_MASTER | NTF_SELF)) {
4453                 NL_SET_ERR_MSG(extack, "Invalid flags in header for fdb get request");
4454                 return -EINVAL;
4455         }
4456
4457         err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct ndmsg), tb,
4458                                             NDA_MAX, nda_policy, extack);
4459         if (err < 0)
4460                 return err;
4461
4462         *ndm_flags = ndm->ndm_flags;
4463         *brport_idx = ndm->ndm_ifindex;
4464         for (i = 0; i <= NDA_MAX; ++i) {
4465                 if (!tb[i])
4466                         continue;
4467
4468                 switch (i) {
4469                 case NDA_MASTER:
4470                         *br_idx = nla_get_u32(tb[i]);
4471                         break;
4472                 case NDA_LLADDR:
4473                         if (nla_len(tb[i]) != ETH_ALEN) {
4474                                 NL_SET_ERR_MSG(extack, "Invalid address in fdb get request");
4475                                 return -EINVAL;
4476                         }
4477                         *addr = nla_data(tb[i]);
4478                         break;
4479                 case NDA_VLAN:
4480                         err = fdb_vid_parse(tb[i], vid, extack);
4481                         if (err)
4482                                 return err;
4483                         break;
4484                 case NDA_VNI:
4485                         break;
4486                 default:
4487                         NL_SET_ERR_MSG(extack, "Unsupported attribute in fdb get request");
4488                         return -EINVAL;
4489                 }
4490         }
4491
4492         return 0;
4493 }
4494
4495 static int rtnl_fdb_get(struct sk_buff *in_skb, struct nlmsghdr *nlh,
4496                         struct netlink_ext_ack *extack)
4497 {
4498         struct net_device *dev = NULL, *br_dev = NULL;
4499         const struct net_device_ops *ops = NULL;
4500         struct net *net = sock_net(in_skb->sk);
4501         struct nlattr *tb[NDA_MAX + 1];
4502         struct sk_buff *skb;
4503         int brport_idx = 0;
4504         u8 ndm_flags = 0;
4505         int br_idx = 0;
4506         u8 *addr = NULL;
4507         u16 vid = 0;
4508         int err;
4509
4510         err = valid_fdb_get_strict(nlh, tb, &ndm_flags, &br_idx,
4511                                    &brport_idx, &addr, &vid, extack);
4512         if (err < 0)
4513                 return err;
4514
4515         if (!addr) {
4516                 NL_SET_ERR_MSG(extack, "Missing lookup address for fdb get request");
4517                 return -EINVAL;
4518         }
4519
4520         if (brport_idx) {
4521                 dev = __dev_get_by_index(net, brport_idx);
4522                 if (!dev) {
4523                         NL_SET_ERR_MSG(extack, "Unknown device ifindex");
4524                         return -ENODEV;
4525                 }
4526         }
4527
4528         if (br_idx) {
4529                 if (dev) {
4530                         NL_SET_ERR_MSG(extack, "Master and device are mutually exclusive");
4531                         return -EINVAL;
4532                 }
4533
4534                 br_dev = __dev_get_by_index(net, br_idx);
4535                 if (!br_dev) {
4536                         NL_SET_ERR_MSG(extack, "Invalid master ifindex");
4537                         return -EINVAL;
4538                 }
4539                 ops = br_dev->netdev_ops;
4540         }
4541
4542         if (dev) {
4543                 if (!ndm_flags || (ndm_flags & NTF_MASTER)) {
4544                         if (!netif_is_bridge_port(dev)) {
4545                                 NL_SET_ERR_MSG(extack, "Device is not a bridge port");
4546                                 return -EINVAL;
4547                         }
4548                         br_dev = netdev_master_upper_dev_get(dev);
4549                         if (!br_dev) {
4550                                 NL_SET_ERR_MSG(extack, "Master of device not found");
4551                                 return -EINVAL;
4552                         }
4553                         ops = br_dev->netdev_ops;
4554                 } else {
4555                         if (!(ndm_flags & NTF_SELF)) {
4556                                 NL_SET_ERR_MSG(extack, "Missing NTF_SELF");
4557                                 return -EINVAL;
4558                         }
4559                         ops = dev->netdev_ops;
4560                 }
4561         }
4562
4563         if (!br_dev && !dev) {
4564                 NL_SET_ERR_MSG(extack, "No device specified");
4565                 return -ENODEV;
4566         }
4567
4568         if (!ops || !ops->ndo_fdb_get) {
4569                 NL_SET_ERR_MSG(extack, "Fdb get operation not supported by device");
4570                 return -EOPNOTSUPP;
4571         }
4572
4573         skb = nlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL);
4574         if (!skb)
4575                 return -ENOBUFS;
4576
4577         if (br_dev)
4578                 dev = br_dev;
4579         err = ops->ndo_fdb_get(skb, tb, dev, addr, vid,
4580                                NETLINK_CB(in_skb).portid,
4581                                nlh->nlmsg_seq, extack);
4582         if (err)
4583                 goto out;
4584
4585         return rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
4586 out:
4587         kfree_skb(skb);
4588         return err;
4589 }
4590
4591 static int brport_nla_put_flag(struct sk_buff *skb, u32 flags, u32 mask,
4592                                unsigned int attrnum, unsigned int flag)
4593 {
4594         if (mask & flag)
4595                 return nla_put_u8(skb, attrnum, !!(flags & flag));
4596         return 0;
4597 }
4598
4599 int ndo_dflt_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
4600                             struct net_device *dev, u16 mode,
4601                             u32 flags, u32 mask, int nlflags,
4602                             u32 filter_mask,
4603                             int (*vlan_fill)(struct sk_buff *skb,
4604                                              struct net_device *dev,
4605                                              u32 filter_mask))
4606 {
4607         struct nlmsghdr *nlh;
4608         struct ifinfomsg *ifm;
4609         struct nlattr *br_afspec;
4610         struct nlattr *protinfo;
4611         u8 operstate = netif_running(dev) ? dev->operstate : IF_OPER_DOWN;
4612         struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4613         int err = 0;
4614
4615         nlh = nlmsg_put(skb, pid, seq, RTM_NEWLINK, sizeof(*ifm), nlflags);
4616         if (nlh == NULL)
4617                 return -EMSGSIZE;
4618
4619         ifm = nlmsg_data(nlh);
4620         ifm->ifi_family = AF_BRIDGE;
4621         ifm->__ifi_pad = 0;
4622         ifm->ifi_type = dev->type;
4623         ifm->ifi_index = dev->ifindex;
4624         ifm->ifi_flags = dev_get_flags(dev);
4625         ifm->ifi_change = 0;
4626
4627
4628         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4629             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4630             nla_put_u8(skb, IFLA_OPERSTATE, operstate) ||
4631             (br_dev &&
4632              nla_put_u32(skb, IFLA_MASTER, br_dev->ifindex)) ||
4633             (dev->addr_len &&
4634              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4635             (dev->ifindex != dev_get_iflink(dev) &&
4636              nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))))
4637                 goto nla_put_failure;
4638
4639         br_afspec = nla_nest_start_noflag(skb, IFLA_AF_SPEC);
4640         if (!br_afspec)
4641                 goto nla_put_failure;
4642
4643         if (nla_put_u16(skb, IFLA_BRIDGE_FLAGS, BRIDGE_FLAGS_SELF)) {
4644                 nla_nest_cancel(skb, br_afspec);
4645                 goto nla_put_failure;
4646         }
4647
4648         if (mode != BRIDGE_MODE_UNDEF) {
4649                 if (nla_put_u16(skb, IFLA_BRIDGE_MODE, mode)) {
4650                         nla_nest_cancel(skb, br_afspec);
4651                         goto nla_put_failure;
4652                 }
4653         }
4654         if (vlan_fill) {
4655                 err = vlan_fill(skb, dev, filter_mask);
4656                 if (err) {
4657                         nla_nest_cancel(skb, br_afspec);
4658                         goto nla_put_failure;
4659                 }
4660         }
4661         nla_nest_end(skb, br_afspec);
4662
4663         protinfo = nla_nest_start(skb, IFLA_PROTINFO);
4664         if (!protinfo)
4665                 goto nla_put_failure;
4666
4667         if (brport_nla_put_flag(skb, flags, mask,
4668                                 IFLA_BRPORT_MODE, BR_HAIRPIN_MODE) ||
4669             brport_nla_put_flag(skb, flags, mask,
4670                                 IFLA_BRPORT_GUARD, BR_BPDU_GUARD) ||
4671             brport_nla_put_flag(skb, flags, mask,
4672                                 IFLA_BRPORT_FAST_LEAVE,
4673                                 BR_MULTICAST_FAST_LEAVE) ||
4674             brport_nla_put_flag(skb, flags, mask,
4675                                 IFLA_BRPORT_PROTECT, BR_ROOT_BLOCK) ||
4676             brport_nla_put_flag(skb, flags, mask,
4677                                 IFLA_BRPORT_LEARNING, BR_LEARNING) ||
4678             brport_nla_put_flag(skb, flags, mask,
4679                                 IFLA_BRPORT_LEARNING_SYNC, BR_LEARNING_SYNC) ||
4680             brport_nla_put_flag(skb, flags, mask,
4681                                 IFLA_BRPORT_UNICAST_FLOOD, BR_FLOOD) ||
4682             brport_nla_put_flag(skb, flags, mask,
4683                                 IFLA_BRPORT_PROXYARP, BR_PROXYARP) ||
4684             brport_nla_put_flag(skb, flags, mask,
4685                                 IFLA_BRPORT_MCAST_FLOOD, BR_MCAST_FLOOD) ||
4686             brport_nla_put_flag(skb, flags, mask,
4687                                 IFLA_BRPORT_BCAST_FLOOD, BR_BCAST_FLOOD)) {
4688                 nla_nest_cancel(skb, protinfo);
4689                 goto nla_put_failure;
4690         }
4691
4692         nla_nest_end(skb, protinfo);
4693
4694         nlmsg_end(skb, nlh);
4695         return 0;
4696 nla_put_failure:
4697         nlmsg_cancel(skb, nlh);
4698         return err ? err : -EMSGSIZE;
4699 }
4700 EXPORT_SYMBOL_GPL(ndo_dflt_bridge_getlink);
4701
4702 static int valid_bridge_getlink_req(const struct nlmsghdr *nlh,
4703                                     bool strict_check, u32 *filter_mask,
4704                                     struct netlink_ext_ack *extack)
4705 {
4706         struct nlattr *tb[IFLA_MAX+1];
4707         int err, i;
4708
4709         if (strict_check) {
4710                 struct ifinfomsg *ifm;
4711
4712                 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) {
4713                         NL_SET_ERR_MSG(extack, "Invalid header for bridge link dump");
4714                         return -EINVAL;
4715                 }
4716
4717                 ifm = nlmsg_data(nlh);
4718                 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags ||
4719                     ifm->ifi_change || ifm->ifi_index) {
4720                         NL_SET_ERR_MSG(extack, "Invalid values in header for bridge link dump request");
4721                         return -EINVAL;
4722                 }
4723
4724                 err = nlmsg_parse_deprecated_strict(nlh,
4725                                                     sizeof(struct ifinfomsg),
4726                                                     tb, IFLA_MAX, ifla_policy,
4727                                                     extack);
4728         } else {
4729                 err = nlmsg_parse_deprecated(nlh, sizeof(struct ifinfomsg),
4730                                              tb, IFLA_MAX, ifla_policy,
4731                                              extack);
4732         }
4733         if (err < 0)
4734                 return err;
4735
4736         /* new attributes should only be added with strict checking */
4737         for (i = 0; i <= IFLA_MAX; ++i) {
4738                 if (!tb[i])
4739                         continue;
4740
4741                 switch (i) {
4742                 case IFLA_EXT_MASK:
4743                         *filter_mask = nla_get_u32(tb[i]);
4744                         break;
4745                 default:
4746                         if (strict_check) {
4747                                 NL_SET_ERR_MSG(extack, "Unsupported attribute in bridge link dump request");
4748                                 return -EINVAL;
4749                         }
4750                 }
4751         }
4752
4753         return 0;
4754 }
4755
4756 static int rtnl_bridge_getlink(struct sk_buff *skb, struct netlink_callback *cb)
4757 {
4758         const struct nlmsghdr *nlh = cb->nlh;
4759         struct net *net = sock_net(skb->sk);
4760         struct net_device *dev;
4761         int idx = 0;
4762         u32 portid = NETLINK_CB(cb->skb).portid;
4763         u32 seq = nlh->nlmsg_seq;
4764         u32 filter_mask = 0;
4765         int err;
4766
4767         err = valid_bridge_getlink_req(nlh, cb->strict_check, &filter_mask,
4768                                        cb->extack);
4769         if (err < 0 && cb->strict_check)
4770                 return err;
4771
4772         rcu_read_lock();
4773         for_each_netdev_rcu(net, dev) {
4774                 const struct net_device_ops *ops = dev->netdev_ops;
4775                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4776
4777                 if (br_dev && br_dev->netdev_ops->ndo_bridge_getlink) {
4778                         if (idx >= cb->args[0]) {
4779                                 err = br_dev->netdev_ops->ndo_bridge_getlink(
4780                                                 skb, portid, seq, dev,
4781                                                 filter_mask, NLM_F_MULTI);
4782                                 if (err < 0 && err != -EOPNOTSUPP) {
4783                                         if (likely(skb->len))
4784                                                 break;
4785
4786                                         goto out_err;
4787                                 }
4788                         }
4789                         idx++;
4790                 }
4791
4792                 if (ops->ndo_bridge_getlink) {
4793                         if (idx >= cb->args[0]) {
4794                                 err = ops->ndo_bridge_getlink(skb, portid,
4795                                                               seq, dev,
4796                                                               filter_mask,
4797                                                               NLM_F_MULTI);
4798                                 if (err < 0 && err != -EOPNOTSUPP) {
4799                                         if (likely(skb->len))
4800                                                 break;
4801
4802                                         goto out_err;
4803                                 }
4804                         }
4805                         idx++;
4806                 }
4807         }
4808         err = skb->len;
4809 out_err:
4810         rcu_read_unlock();
4811         cb->args[0] = idx;
4812
4813         return err;
4814 }
4815
4816 static inline size_t bridge_nlmsg_size(void)
4817 {
4818         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
4819                 + nla_total_size(IFNAMSIZ)      /* IFLA_IFNAME */
4820                 + nla_total_size(MAX_ADDR_LEN)  /* IFLA_ADDRESS */
4821                 + nla_total_size(sizeof(u32))   /* IFLA_MASTER */
4822                 + nla_total_size(sizeof(u32))   /* IFLA_MTU */
4823                 + nla_total_size(sizeof(u32))   /* IFLA_LINK */
4824                 + nla_total_size(sizeof(u32))   /* IFLA_OPERSTATE */
4825                 + nla_total_size(sizeof(u8))    /* IFLA_PROTINFO */
4826                 + nla_total_size(sizeof(struct nlattr)) /* IFLA_AF_SPEC */
4827                 + nla_total_size(sizeof(u16))   /* IFLA_BRIDGE_FLAGS */
4828                 + nla_total_size(sizeof(u16));  /* IFLA_BRIDGE_MODE */
4829 }
4830
4831 static int rtnl_bridge_notify(struct net_device *dev)
4832 {
4833         struct net *net = dev_net(dev);
4834         struct sk_buff *skb;
4835         int err = -EOPNOTSUPP;
4836
4837         if (!dev->netdev_ops->ndo_bridge_getlink)
4838                 return 0;
4839
4840         skb = nlmsg_new(bridge_nlmsg_size(), GFP_ATOMIC);
4841         if (!skb) {
4842                 err = -ENOMEM;
4843                 goto errout;
4844         }
4845
4846         err = dev->netdev_ops->ndo_bridge_getlink(skb, 0, 0, dev, 0, 0);
4847         if (err < 0)
4848                 goto errout;
4849
4850         /* Notification info is only filled for bridge ports, not the bridge
4851          * device itself. Therefore, a zero notification length is valid and
4852          * should not result in an error.
4853          */
4854         if (!skb->len)
4855                 goto errout;
4856
4857         rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
4858         return 0;
4859 errout:
4860         WARN_ON(err == -EMSGSIZE);
4861         kfree_skb(skb);
4862         if (err)
4863                 rtnl_set_sk_err(net, RTNLGRP_LINK, err);
4864         return err;
4865 }
4866
4867 static int rtnl_bridge_setlink(struct sk_buff *skb, struct nlmsghdr *nlh,
4868                                struct netlink_ext_ack *extack)
4869 {
4870         struct net *net = sock_net(skb->sk);
4871         struct ifinfomsg *ifm;
4872         struct net_device *dev;
4873         struct nlattr *br_spec, *attr = NULL;
4874         int rem, err = -EOPNOTSUPP;
4875         u16 flags = 0;
4876         bool have_flags = false;
4877
4878         if (nlmsg_len(nlh) < sizeof(*ifm))
4879                 return -EINVAL;
4880
4881         ifm = nlmsg_data(nlh);
4882         if (ifm->ifi_family != AF_BRIDGE)
4883                 return -EPFNOSUPPORT;
4884
4885         dev = __dev_get_by_index(net, ifm->ifi_index);
4886         if (!dev) {
4887                 NL_SET_ERR_MSG(extack, "unknown ifindex");
4888                 return -ENODEV;
4889         }
4890
4891         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
4892         if (br_spec) {
4893                 nla_for_each_nested(attr, br_spec, rem) {
4894                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
4895                                 if (nla_len(attr) < sizeof(flags))
4896                                         return -EINVAL;
4897
4898                                 have_flags = true;
4899                                 flags = nla_get_u16(attr);
4900                                 break;
4901                         }
4902                 }
4903         }
4904
4905         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
4906                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4907
4908                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_setlink) {
4909                         err = -EOPNOTSUPP;
4910                         goto out;
4911                 }
4912
4913                 err = br_dev->netdev_ops->ndo_bridge_setlink(dev, nlh, flags,
4914                                                              extack);
4915                 if (err)
4916                         goto out;
4917
4918                 flags &= ~BRIDGE_FLAGS_MASTER;
4919         }
4920
4921         if ((flags & BRIDGE_FLAGS_SELF)) {
4922                 if (!dev->netdev_ops->ndo_bridge_setlink)
4923                         err = -EOPNOTSUPP;
4924                 else
4925                         err = dev->netdev_ops->ndo_bridge_setlink(dev, nlh,
4926                                                                   flags,
4927                                                                   extack);
4928                 if (!err) {
4929                         flags &= ~BRIDGE_FLAGS_SELF;
4930
4931                         /* Generate event to notify upper layer of bridge
4932                          * change
4933                          */
4934                         err = rtnl_bridge_notify(dev);
4935                 }
4936         }
4937
4938         if (have_flags)
4939                 memcpy(nla_data(attr), &flags, sizeof(flags));
4940 out:
4941         return err;
4942 }
4943
4944 static int rtnl_bridge_dellink(struct sk_buff *skb, struct nlmsghdr *nlh,
4945                                struct netlink_ext_ack *extack)
4946 {
4947         struct net *net = sock_net(skb->sk);
4948         struct ifinfomsg *ifm;
4949         struct net_device *dev;
4950         struct nlattr *br_spec, *attr = NULL;
4951         int rem, err = -EOPNOTSUPP;
4952         u16 flags = 0;
4953         bool have_flags = false;
4954
4955         if (nlmsg_len(nlh) < sizeof(*ifm))
4956                 return -EINVAL;
4957
4958         ifm = nlmsg_data(nlh);
4959         if (ifm->ifi_family != AF_BRIDGE)
4960                 return -EPFNOSUPPORT;
4961
4962         dev = __dev_get_by_index(net, ifm->ifi_index);
4963         if (!dev) {
4964                 NL_SET_ERR_MSG(extack, "unknown ifindex");
4965                 return -ENODEV;
4966         }
4967
4968         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
4969         if (br_spec) {
4970                 nla_for_each_nested(attr, br_spec, rem) {
4971                         if (nla_type(attr) == IFLA_BRIDGE_FLAGS) {
4972                                 if (nla_len(attr) < sizeof(flags))
4973                                         return -EINVAL;
4974
4975                                 have_flags = true;
4976                                 flags = nla_get_u16(attr);
4977                                 break;
4978                         }
4979                 }
4980         }
4981
4982         if (!flags || (flags & BRIDGE_FLAGS_MASTER)) {
4983                 struct net_device *br_dev = netdev_master_upper_dev_get(dev);
4984
4985                 if (!br_dev || !br_dev->netdev_ops->ndo_bridge_dellink) {
4986                         err = -EOPNOTSUPP;
4987                         goto out;
4988                 }
4989
4990                 err = br_dev->netdev_ops->ndo_bridge_dellink(dev, nlh, flags);
4991                 if (err)
4992                         goto out;
4993
4994                 flags &= ~BRIDGE_FLAGS_MASTER;
4995         }
4996
4997         if ((flags & BRIDGE_FLAGS_SELF)) {
4998                 if (!dev->netdev_ops->ndo_bridge_dellink)
4999                         err = -EOPNOTSUPP;
5000                 else
5001                         err = dev->netdev_ops->ndo_bridge_dellink(dev, nlh,
5002                                                                   flags);
5003
5004                 if (!err) {
5005                         flags &= ~BRIDGE_FLAGS_SELF;
5006
5007                         /* Generate event to notify upper layer of bridge
5008                          * change
5009                          */
5010                         err = rtnl_bridge_notify(dev);
5011                 }
5012         }
5013
5014         if (have_flags)
5015                 memcpy(nla_data(attr), &flags, sizeof(flags));
5016 out:
5017         return err;
5018 }
5019
5020 static bool stats_attr_valid(unsigned int mask, int attrid, int idxattr)
5021 {
5022         return (mask & IFLA_STATS_FILTER_BIT(attrid)) &&
5023                (!idxattr || idxattr == attrid);
5024 }
5025
5026 #define IFLA_OFFLOAD_XSTATS_FIRST (IFLA_OFFLOAD_XSTATS_UNSPEC + 1)
5027 static int rtnl_get_offload_stats_attr_size(int attr_id)
5028 {
5029         switch (attr_id) {
5030         case IFLA_OFFLOAD_XSTATS_CPU_HIT:
5031                 return sizeof(struct rtnl_link_stats64);
5032         }
5033
5034         return 0;
5035 }
5036
5037 static int rtnl_get_offload_stats(struct sk_buff *skb, struct net_device *dev,
5038                                   int *prividx)
5039 {
5040         struct nlattr *attr = NULL;
5041         int attr_id, size;
5042         void *attr_data;
5043         int err;
5044
5045         if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
5046               dev->netdev_ops->ndo_get_offload_stats))
5047                 return -ENODATA;
5048
5049         for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
5050              attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
5051                 if (attr_id < *prividx)
5052                         continue;
5053
5054                 size = rtnl_get_offload_stats_attr_size(attr_id);
5055                 if (!size)
5056                         continue;
5057
5058                 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
5059                         continue;
5060
5061                 attr = nla_reserve_64bit(skb, attr_id, size,
5062                                          IFLA_OFFLOAD_XSTATS_UNSPEC);
5063                 if (!attr)
5064                         goto nla_put_failure;
5065
5066                 attr_data = nla_data(attr);
5067                 memset(attr_data, 0, size);
5068                 err = dev->netdev_ops->ndo_get_offload_stats(attr_id, dev,
5069                                                              attr_data);
5070                 if (err)
5071                         goto get_offload_stats_failure;
5072         }
5073
5074         if (!attr)
5075                 return -ENODATA;
5076
5077         *prividx = 0;
5078         return 0;
5079
5080 nla_put_failure:
5081         err = -EMSGSIZE;
5082 get_offload_stats_failure:
5083         *prividx = attr_id;
5084         return err;
5085 }
5086
5087 static int rtnl_get_offload_stats_size(const struct net_device *dev)
5088 {
5089         int nla_size = 0;
5090         int attr_id;
5091         int size;
5092
5093         if (!(dev->netdev_ops && dev->netdev_ops->ndo_has_offload_stats &&
5094               dev->netdev_ops->ndo_get_offload_stats))
5095                 return 0;
5096
5097         for (attr_id = IFLA_OFFLOAD_XSTATS_FIRST;
5098              attr_id <= IFLA_OFFLOAD_XSTATS_MAX; attr_id++) {
5099                 if (!dev->netdev_ops->ndo_has_offload_stats(dev, attr_id))
5100                         continue;
5101                 size = rtnl_get_offload_stats_attr_size(attr_id);
5102                 nla_size += nla_total_size_64bit(size);
5103         }
5104
5105         if (nla_size != 0)
5106                 nla_size += nla_total_size(0);
5107
5108         return nla_size;
5109 }
5110
5111 static int rtnl_fill_statsinfo(struct sk_buff *skb, struct net_device *dev,
5112                                int type, u32 pid, u32 seq, u32 change,
5113                                unsigned int flags, unsigned int filter_mask,
5114                                int *idxattr, int *prividx)
5115 {
5116         struct if_stats_msg *ifsm;
5117         struct nlmsghdr *nlh;
5118         struct nlattr *attr;
5119         int s_prividx = *prividx;
5120         int err;
5121
5122         ASSERT_RTNL();
5123
5124         nlh = nlmsg_put(skb, pid, seq, type, sizeof(*ifsm), flags);
5125         if (!nlh)
5126                 return -EMSGSIZE;
5127
5128         ifsm = nlmsg_data(nlh);
5129         ifsm->family = PF_UNSPEC;
5130         ifsm->pad1 = 0;
5131         ifsm->pad2 = 0;
5132         ifsm->ifindex = dev->ifindex;
5133         ifsm->filter_mask = filter_mask;
5134
5135         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, *idxattr)) {
5136                 struct rtnl_link_stats64 *sp;
5137
5138                 attr = nla_reserve_64bit(skb, IFLA_STATS_LINK_64,
5139                                          sizeof(struct rtnl_link_stats64),
5140                                          IFLA_STATS_UNSPEC);
5141                 if (!attr)
5142                         goto nla_put_failure;
5143
5144                 sp = nla_data(attr);
5145                 dev_get_stats(dev, sp);
5146         }
5147
5148         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, *idxattr)) {
5149                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
5150
5151                 if (ops && ops->fill_linkxstats) {
5152                         *idxattr = IFLA_STATS_LINK_XSTATS;
5153                         attr = nla_nest_start_noflag(skb,
5154                                                      IFLA_STATS_LINK_XSTATS);
5155                         if (!attr)
5156                                 goto nla_put_failure;
5157
5158                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
5159                         nla_nest_end(skb, attr);
5160                         if (err)
5161                                 goto nla_put_failure;
5162                         *idxattr = 0;
5163                 }
5164         }
5165
5166         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE,
5167                              *idxattr)) {
5168                 const struct rtnl_link_ops *ops = NULL;
5169                 const struct net_device *master;
5170
5171                 master = netdev_master_upper_dev_get(dev);
5172                 if (master)
5173                         ops = master->rtnl_link_ops;
5174                 if (ops && ops->fill_linkxstats) {
5175                         *idxattr = IFLA_STATS_LINK_XSTATS_SLAVE;
5176                         attr = nla_nest_start_noflag(skb,
5177                                                      IFLA_STATS_LINK_XSTATS_SLAVE);
5178                         if (!attr)
5179                                 goto nla_put_failure;
5180
5181                         err = ops->fill_linkxstats(skb, dev, prividx, *idxattr);
5182                         nla_nest_end(skb, attr);
5183                         if (err)
5184                                 goto nla_put_failure;
5185                         *idxattr = 0;
5186                 }
5187         }
5188
5189         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS,
5190                              *idxattr)) {
5191                 *idxattr = IFLA_STATS_LINK_OFFLOAD_XSTATS;
5192                 attr = nla_nest_start_noflag(skb,
5193                                              IFLA_STATS_LINK_OFFLOAD_XSTATS);
5194                 if (!attr)
5195                         goto nla_put_failure;
5196
5197                 err = rtnl_get_offload_stats(skb, dev, prividx);
5198                 if (err == -ENODATA)
5199                         nla_nest_cancel(skb, attr);
5200                 else
5201                         nla_nest_end(skb, attr);
5202
5203                 if (err && err != -ENODATA)
5204                         goto nla_put_failure;
5205                 *idxattr = 0;
5206         }
5207
5208         if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, *idxattr)) {
5209                 struct rtnl_af_ops *af_ops;
5210
5211                 *idxattr = IFLA_STATS_AF_SPEC;
5212                 attr = nla_nest_start_noflag(skb, IFLA_STATS_AF_SPEC);
5213                 if (!attr)
5214                         goto nla_put_failure;
5215
5216                 rcu_read_lock();
5217                 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
5218                         if (af_ops->fill_stats_af) {
5219                                 struct nlattr *af;
5220                                 int err;
5221
5222                                 af = nla_nest_start_noflag(skb,
5223                                                            af_ops->family);
5224                                 if (!af) {
5225                                         rcu_read_unlock();
5226                                         goto nla_put_failure;
5227                                 }
5228                                 err = af_ops->fill_stats_af(skb, dev);
5229
5230                                 if (err == -ENODATA) {
5231                                         nla_nest_cancel(skb, af);
5232                                 } else if (err < 0) {
5233                                         rcu_read_unlock();
5234                                         goto nla_put_failure;
5235                                 }
5236
5237                                 nla_nest_end(skb, af);
5238                         }
5239                 }
5240                 rcu_read_unlock();
5241
5242                 nla_nest_end(skb, attr);
5243
5244                 *idxattr = 0;
5245         }
5246
5247         nlmsg_end(skb, nlh);
5248
5249         return 0;
5250
5251 nla_put_failure:
5252         /* not a multi message or no progress mean a real error */
5253         if (!(flags & NLM_F_MULTI) || s_prividx == *prividx)
5254                 nlmsg_cancel(skb, nlh);
5255         else
5256                 nlmsg_end(skb, nlh);
5257
5258         return -EMSGSIZE;
5259 }
5260
5261 static size_t if_nlmsg_stats_size(const struct net_device *dev,
5262                                   u32 filter_mask)
5263 {
5264         size_t size = NLMSG_ALIGN(sizeof(struct if_stats_msg));
5265
5266         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_64, 0))
5267                 size += nla_total_size_64bit(sizeof(struct rtnl_link_stats64));
5268
5269         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS, 0)) {
5270                 const struct rtnl_link_ops *ops = dev->rtnl_link_ops;
5271                 int attr = IFLA_STATS_LINK_XSTATS;
5272
5273                 if (ops && ops->get_linkxstats_size) {
5274                         size += nla_total_size(ops->get_linkxstats_size(dev,
5275                                                                         attr));
5276                         /* for IFLA_STATS_LINK_XSTATS */
5277                         size += nla_total_size(0);
5278                 }
5279         }
5280
5281         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_XSTATS_SLAVE, 0)) {
5282                 struct net_device *_dev = (struct net_device *)dev;
5283                 const struct rtnl_link_ops *ops = NULL;
5284                 const struct net_device *master;
5285
5286                 /* netdev_master_upper_dev_get can't take const */
5287                 master = netdev_master_upper_dev_get(_dev);
5288                 if (master)
5289                         ops = master->rtnl_link_ops;
5290                 if (ops && ops->get_linkxstats_size) {
5291                         int attr = IFLA_STATS_LINK_XSTATS_SLAVE;
5292
5293                         size += nla_total_size(ops->get_linkxstats_size(dev,
5294                                                                         attr));
5295                         /* for IFLA_STATS_LINK_XSTATS_SLAVE */
5296                         size += nla_total_size(0);
5297                 }
5298         }
5299
5300         if (stats_attr_valid(filter_mask, IFLA_STATS_LINK_OFFLOAD_XSTATS, 0))
5301                 size += rtnl_get_offload_stats_size(dev);
5302
5303         if (stats_attr_valid(filter_mask, IFLA_STATS_AF_SPEC, 0)) {
5304                 struct rtnl_af_ops *af_ops;
5305
5306                 /* for IFLA_STATS_AF_SPEC */
5307                 size += nla_total_size(0);
5308
5309                 rcu_read_lock();
5310                 list_for_each_entry_rcu(af_ops, &rtnl_af_ops, list) {
5311                         if (af_ops->get_stats_af_size) {
5312                                 size += nla_total_size(
5313                                         af_ops->get_stats_af_size(dev));
5314
5315                                 /* for AF_* */
5316                                 size += nla_total_size(0);
5317                         }
5318                 }
5319                 rcu_read_unlock();
5320         }
5321
5322         return size;
5323 }
5324
5325 static int rtnl_valid_stats_req(const struct nlmsghdr *nlh, bool strict_check,
5326                                 bool is_dump, struct netlink_ext_ack *extack)
5327 {
5328         struct if_stats_msg *ifsm;
5329
5330         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifsm))) {
5331                 NL_SET_ERR_MSG(extack, "Invalid header for stats dump");
5332                 return -EINVAL;
5333         }
5334
5335         if (!strict_check)
5336                 return 0;
5337
5338         ifsm = nlmsg_data(nlh);
5339
5340         /* only requests using strict checks can pass data to influence
5341          * the dump. The legacy exception is filter_mask.
5342          */
5343         if (ifsm->pad1 || ifsm->pad2 || (is_dump && ifsm->ifindex)) {
5344                 NL_SET_ERR_MSG(extack, "Invalid values in header for stats dump request");
5345                 return -EINVAL;
5346         }
5347         if (nlmsg_attrlen(nlh, sizeof(*ifsm))) {
5348                 NL_SET_ERR_MSG(extack, "Invalid attributes after stats header");
5349                 return -EINVAL;
5350         }
5351         if (ifsm->filter_mask >= IFLA_STATS_FILTER_BIT(IFLA_STATS_MAX + 1)) {
5352                 NL_SET_ERR_MSG(extack, "Invalid stats requested through filter mask");
5353                 return -EINVAL;
5354         }
5355
5356         return 0;
5357 }
5358
5359 static int rtnl_stats_get(struct sk_buff *skb, struct nlmsghdr *nlh,
5360                           struct netlink_ext_ack *extack)
5361 {
5362         struct net *net = sock_net(skb->sk);
5363         struct net_device *dev = NULL;
5364         int idxattr = 0, prividx = 0;
5365         struct if_stats_msg *ifsm;
5366         struct sk_buff *nskb;
5367         u32 filter_mask;
5368         int err;
5369
5370         err = rtnl_valid_stats_req(nlh, netlink_strict_get_check(skb),
5371                                    false, extack);
5372         if (err)
5373                 return err;
5374
5375         ifsm = nlmsg_data(nlh);
5376         if (ifsm->ifindex > 0)
5377                 dev = __dev_get_by_index(net, ifsm->ifindex);
5378         else
5379                 return -EINVAL;
5380
5381         if (!dev)
5382                 return -ENODEV;
5383
5384         filter_mask = ifsm->filter_mask;
5385         if (!filter_mask)
5386                 return -EINVAL;
5387
5388         nskb = nlmsg_new(if_nlmsg_stats_size(dev, filter_mask), GFP_KERNEL);
5389         if (!nskb)
5390                 return -ENOBUFS;
5391
5392         err = rtnl_fill_statsinfo(nskb, dev, RTM_NEWSTATS,
5393                                   NETLINK_CB(skb).portid, nlh->nlmsg_seq, 0,
5394                                   0, filter_mask, &idxattr, &prividx);
5395         if (err < 0) {
5396                 /* -EMSGSIZE implies BUG in if_nlmsg_stats_size */
5397                 WARN_ON(err == -EMSGSIZE);
5398                 kfree_skb(nskb);
5399         } else {
5400                 err = rtnl_unicast(nskb, net, NETLINK_CB(skb).portid);
5401         }
5402
5403         return err;
5404 }
5405
5406 static int rtnl_stats_dump(struct sk_buff *skb, struct netlink_callback *cb)
5407 {
5408         struct netlink_ext_ack *extack = cb->extack;
5409         int h, s_h, err, s_idx, s_idxattr, s_prividx;
5410         struct net *net = sock_net(skb->sk);
5411         unsigned int flags = NLM_F_MULTI;
5412         struct if_stats_msg *ifsm;
5413         struct hlist_head *head;
5414         struct net_device *dev;
5415         u32 filter_mask = 0;
5416         int idx = 0;
5417
5418         s_h = cb->args[0];
5419         s_idx = cb->args[1];
5420         s_idxattr = cb->args[2];
5421         s_prividx = cb->args[3];
5422
5423         cb->seq = net->dev_base_seq;
5424
5425         err = rtnl_valid_stats_req(cb->nlh, cb->strict_check, true, extack);
5426         if (err)
5427                 return err;
5428
5429         ifsm = nlmsg_data(cb->nlh);
5430         filter_mask = ifsm->filter_mask;
5431         if (!filter_mask) {
5432                 NL_SET_ERR_MSG(extack, "Filter mask must be set for stats dump");
5433                 return -EINVAL;
5434         }
5435
5436         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
5437                 idx = 0;
5438                 head = &net->dev_index_head[h];
5439                 hlist_for_each_entry(dev, head, index_hlist) {
5440                         if (idx < s_idx)
5441                                 goto cont;
5442                         err = rtnl_fill_statsinfo(skb, dev, RTM_NEWSTATS,
5443                                                   NETLINK_CB(cb->skb).portid,
5444                                                   cb->nlh->nlmsg_seq, 0,
5445                                                   flags, filter_mask,
5446                                                   &s_idxattr, &s_prividx);
5447                         /* If we ran out of room on the first message,
5448                          * we're in trouble
5449                          */
5450                         WARN_ON((err == -EMSGSIZE) && (skb->len == 0));
5451
5452                         if (err < 0)
5453                                 goto out;
5454                         s_prividx = 0;
5455                         s_idxattr = 0;
5456                         nl_dump_check_consistent(cb, nlmsg_hdr(skb));
5457 cont:
5458                         idx++;
5459                 }
5460         }
5461 out:
5462         cb->args[3] = s_prividx;
5463         cb->args[2] = s_idxattr;
5464         cb->args[1] = idx;
5465         cb->args[0] = h;
5466
5467         return skb->len;
5468 }
5469
5470 /* Process one rtnetlink message. */
5471
5472 static int rtnetlink_rcv_msg(struct sk_buff *skb, struct nlmsghdr *nlh,
5473                              struct netlink_ext_ack *extack)
5474 {
5475         struct net *net = sock_net(skb->sk);
5476         struct rtnl_link *link;
5477         struct module *owner;
5478         int err = -EOPNOTSUPP;
5479         rtnl_doit_func doit;
5480         unsigned int flags;
5481         int kind;
5482         int family;
5483         int type;
5484
5485         type = nlh->nlmsg_type;
5486         if (type > RTM_MAX)
5487                 return -EOPNOTSUPP;
5488
5489         type -= RTM_BASE;
5490
5491         /* All the messages must have at least 1 byte length */
5492         if (nlmsg_len(nlh) < sizeof(struct rtgenmsg))
5493                 return 0;
5494
5495         family = ((struct rtgenmsg *)nlmsg_data(nlh))->rtgen_family;
5496         kind = type&3;
5497
5498         if (kind != 2 && !netlink_net_capable(skb, CAP_NET_ADMIN))
5499                 return -EPERM;
5500
5501         rcu_read_lock();
5502         if (kind == 2 && nlh->nlmsg_flags&NLM_F_DUMP) {
5503                 struct sock *rtnl;
5504                 rtnl_dumpit_func dumpit;
5505                 u32 min_dump_alloc = 0;
5506
5507                 link = rtnl_get_link(family, type);
5508                 if (!link || !link->dumpit) {
5509                         family = PF_UNSPEC;
5510                         link = rtnl_get_link(family, type);
5511                         if (!link || !link->dumpit)
5512                                 goto err_unlock;
5513                 }
5514                 owner = link->owner;
5515                 dumpit = link->dumpit;
5516
5517                 if (type == RTM_GETLINK - RTM_BASE)
5518                         min_dump_alloc = rtnl_calcit(skb, nlh);
5519
5520                 err = 0;
5521                 /* need to do this before rcu_read_unlock() */
5522                 if (!try_module_get(owner))
5523                         err = -EPROTONOSUPPORT;
5524
5525                 rcu_read_unlock();
5526
5527                 rtnl = net->rtnl;
5528                 if (err == 0) {
5529                         struct netlink_dump_control c = {
5530                                 .dump           = dumpit,
5531                                 .min_dump_alloc = min_dump_alloc,
5532                                 .module         = owner,
5533                         };
5534                         err = netlink_dump_start(rtnl, skb, nlh, &c);
5535                         /* netlink_dump_start() will keep a reference on
5536                          * module if dump is still in progress.
5537                          */
5538                         module_put(owner);
5539                 }
5540                 return err;
5541         }
5542
5543         link = rtnl_get_link(family, type);
5544         if (!link || !link->doit) {
5545                 family = PF_UNSPEC;
5546                 link = rtnl_get_link(PF_UNSPEC, type);
5547                 if (!link || !link->doit)
5548                         goto out_unlock;
5549         }
5550
5551         owner = link->owner;
5552         if (!try_module_get(owner)) {
5553                 err = -EPROTONOSUPPORT;
5554                 goto out_unlock;
5555         }
5556
5557         flags = link->flags;
5558         if (flags & RTNL_FLAG_DOIT_UNLOCKED) {
5559                 doit = link->doit;
5560                 rcu_read_unlock();
5561                 if (doit)
5562                         err = doit(skb, nlh, extack);
5563                 module_put(owner);
5564                 return err;
5565         }
5566         rcu_read_unlock();
5567
5568         rtnl_lock();
5569         link = rtnl_get_link(family, type);
5570         if (link && link->doit)
5571                 err = link->doit(skb, nlh, extack);
5572         rtnl_unlock();
5573
5574         module_put(owner);
5575
5576         return err;
5577
5578 out_unlock:
5579         rcu_read_unlock();
5580         return err;
5581
5582 err_unlock:
5583         rcu_read_unlock();
5584         return -EOPNOTSUPP;
5585 }
5586
5587 static void rtnetlink_rcv(struct sk_buff *skb)
5588 {
5589         netlink_rcv_skb(skb, &rtnetlink_rcv_msg);
5590 }
5591
5592 static int rtnetlink_bind(struct net *net, int group)
5593 {
5594         switch (group) {
5595         case RTNLGRP_IPV4_MROUTE_R:
5596         case RTNLGRP_IPV6_MROUTE_R:
5597                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
5598                         return -EPERM;
5599                 break;
5600         }
5601         return 0;
5602 }
5603
5604 static int rtnetlink_event(struct notifier_block *this, unsigned long event, void *ptr)
5605 {
5606         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
5607
5608         switch (event) {
5609         case NETDEV_REBOOT:
5610         case NETDEV_CHANGEMTU:
5611         case NETDEV_CHANGEADDR:
5612         case NETDEV_CHANGENAME:
5613         case NETDEV_FEAT_CHANGE:
5614         case NETDEV_BONDING_FAILOVER:
5615         case NETDEV_POST_TYPE_CHANGE:
5616         case NETDEV_NOTIFY_PEERS:
5617         case NETDEV_CHANGEUPPER:
5618         case NETDEV_RESEND_IGMP:
5619         case NETDEV_CHANGEINFODATA:
5620         case NETDEV_CHANGELOWERSTATE:
5621         case NETDEV_CHANGE_TX_QUEUE_LEN:
5622                 rtmsg_ifinfo_event(RTM_NEWLINK, dev, 0, rtnl_get_event(event),
5623                                    GFP_KERNEL, NULL, 0);
5624                 break;
5625         default:
5626                 break;
5627         }
5628         return NOTIFY_DONE;
5629 }
5630
5631 static struct notifier_block rtnetlink_dev_notifier = {
5632         .notifier_call  = rtnetlink_event,
5633 };
5634
5635
5636 static int __net_init rtnetlink_net_init(struct net *net)
5637 {
5638         struct sock *sk;
5639         struct netlink_kernel_cfg cfg = {
5640                 .groups         = RTNLGRP_MAX,
5641                 .input          = rtnetlink_rcv,
5642                 .cb_mutex       = &rtnl_mutex,
5643                 .flags          = NL_CFG_F_NONROOT_RECV,
5644                 .bind           = rtnetlink_bind,
5645         };
5646
5647         sk = netlink_kernel_create(net, NETLINK_ROUTE, &cfg);
5648         if (!sk)
5649                 return -ENOMEM;
5650         net->rtnl = sk;
5651         return 0;
5652 }
5653
5654 static void __net_exit rtnetlink_net_exit(struct net *net)
5655 {
5656         netlink_kernel_release(net->rtnl);
5657         net->rtnl = NULL;
5658 }
5659
5660 static struct pernet_operations rtnetlink_net_ops = {
5661         .init = rtnetlink_net_init,
5662         .exit = rtnetlink_net_exit,
5663 };
5664
5665 void __init rtnetlink_init(void)
5666 {
5667         if (register_pernet_subsys(&rtnetlink_net_ops))
5668                 panic("rtnetlink_init: cannot initialize rtnetlink\n");
5669
5670         register_netdevice_notifier(&rtnetlink_dev_notifier);
5671
5672         rtnl_register(PF_UNSPEC, RTM_GETLINK, rtnl_getlink,
5673                       rtnl_dump_ifinfo, 0);
5674         rtnl_register(PF_UNSPEC, RTM_SETLINK, rtnl_setlink, NULL, 0);
5675         rtnl_register(PF_UNSPEC, RTM_NEWLINK, rtnl_newlink, NULL, 0);
5676         rtnl_register(PF_UNSPEC, RTM_DELLINK, rtnl_dellink, NULL, 0);
5677
5678         rtnl_register(PF_UNSPEC, RTM_GETADDR, NULL, rtnl_dump_all, 0);
5679         rtnl_register(PF_UNSPEC, RTM_GETROUTE, NULL, rtnl_dump_all, 0);
5680         rtnl_register(PF_UNSPEC, RTM_GETNETCONF, NULL, rtnl_dump_all, 0);
5681
5682         rtnl_register(PF_UNSPEC, RTM_NEWLINKPROP, rtnl_newlinkprop, NULL, 0);
5683         rtnl_register(PF_UNSPEC, RTM_DELLINKPROP, rtnl_dellinkprop, NULL, 0);
5684
5685         rtnl_register(PF_BRIDGE, RTM_NEWNEIGH, rtnl_fdb_add, NULL, 0);
5686         rtnl_register(PF_BRIDGE, RTM_DELNEIGH, rtnl_fdb_del, NULL, 0);
5687         rtnl_register(PF_BRIDGE, RTM_GETNEIGH, rtnl_fdb_get, rtnl_fdb_dump, 0);
5688
5689         rtnl_register(PF_BRIDGE, RTM_GETLINK, NULL, rtnl_bridge_getlink, 0);
5690         rtnl_register(PF_BRIDGE, RTM_DELLINK, rtnl_bridge_dellink, NULL, 0);
5691         rtnl_register(PF_BRIDGE, RTM_SETLINK, rtnl_bridge_setlink, NULL, 0);
5692
5693         rtnl_register(PF_UNSPEC, RTM_GETSTATS, rtnl_stats_get, rtnl_stats_dump,
5694                       0);
5695 }